survey validity and reliability study · pdf filevalidity%and%reliability%of ......

85
Page | 1 Technical Report: Examination of the Psychometric Properties of the Catholic School Program Effectiveness and Defining Characteristics Surveys Prepared for: National Task Force, National Standards and Benchmarks for Effective Catholic Elementary and Secondary Schools Center for Catholic School Effectiveness, Loyola University Chicago Barbara and Patrick Roche Center for Catholic Education, Boston College AdvancED Author: Scott R. Weaver, Ph.D. AdvancED Consultant July 31, 2012 Acknowledgments: Special thanks are extended to Dr. Eddie Krenson, Dr. Vicki Denmark, and Karmen Gary at AdvancED, for their assistance with the coordination of multiple elements of this project. The surveys examined in this psychometric study were developed by the following national task force with support from AdvancED: Lorraine A. Ozar, Ph.D, Loyola University Chicago (Chair); Susan Ferguson, Ed.D., University of Dayton; Adam Kruekeberg, MBA/MA Pastoral Ministry, Boston College; Kathleen Schwartz, Ed.D., Diocese of Venice, FL; Patricia Weitzel O’Neill, Ph.D., Boston College.

Upload: hoangngoc

Post on 15-Mar-2018

215 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  1      

 

Technical  Report:  Examination  of  the  Psychometric  Properties  of  the    

Catholic  School  Program  Effectiveness  and    Defining  Characteristics  Surveys  

Prepared for: National Task Force, National Standards and Benchmarks for Effective Catholic Elementary and Secondary Schools

Center for Catholic School Effectiveness, Loyola University Chicago

Barbara and Patrick Roche Center for Catholic Education, Boston College

AdvancED

 

Author:  Scott R. Weaver, Ph.D. AdvancED Consultant

 July 31, 2012  

Acknowledgments: Special  thanks  are  extended  to  Dr.  Eddie  Krenson,  Dr.  Vicki  Denmark,  and  Karmen  Gary  at  AdvancED,  for  their  assistance  with  the  coordination  of  multiple  elements  of  this  project.          The  surveys  examined  in  this  psychometric  study  were  developed  by  the  following  national  task  force  with  support  from  AdvancED:    Lorraine  A.  Ozar,  Ph.D,  Loyola  University  Chicago  (Chair);  Susan  Ferguson,  Ed.D.,  University  of  Dayton;    Adam  Kruekeberg,  MBA/MA  Pastoral  Ministry,  Boston  College;  Kathleen  Schwartz,  Ed.D.,  Diocese  of  Venice,  FL;  Patricia  Weitzel-­‐O’Neill,  Ph.D.,  Boston  College.  

Page 2: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  2      

 

 Table  of  Contents  

EXECUTIVE  SUMMARY...............................................................................................................................................................4  SAMPLING  AND  METHODOLOGY ...............................................................................................................................................6  

SAMPLING  OVERVIEW ...........................................................................................................................................................................6  METHODOLOGY ...................................................................................................................................................................................6  

PROGRAM  EFFECTIVENESS  SURVEYS........................................................................................................................................ 10  

DESCRIPTION .....................................................................................................................................................................................10  ADULT  SURVEY ..................................................................................................................................................................................10  

Respondent  Characteristics .......................................................................................................................................................10  Response  Pattern  Summaries....................................................................................................................................................11  Factorial  Validity  and  Reliability................................................................................................................................................12  Summary  and  Recommendations .............................................................................................................................................16  

STUDENT  SURVEY  (GRADES  5-­‐8) ...........................................................................................................................................................19  

Respondent  Characteristics .......................................................................................................................................................19  Response  Pattern  Summaries....................................................................................................................................................19  Factorial  Validity  and  Reliability................................................................................................................................................20  Summary  and  Recommendations .............................................................................................................................................25  

STUDENT  SURVEY  (GRADES  9-­‐12) .........................................................................................................................................................28  

Respondent  Characteristics .......................................................................................................................................................28  Response  Pattern  Summaries....................................................................................................................................................28  Factorial  Validity  and  Reliability................................................................................................................................................29  Summary  and  Recommendations .............................................................................................................................................34  

DEFINING  CHARACTERISTICS  SURVEYS..................................................................................................................................... 36  

DESCRIPTION .....................................................................................................................................................................................36  ADULT  SURVEY ..................................................................................................................................................................................36  

Respondent  Characteristics .......................................................................................................................................................36  Response  Pattern  Summaries....................................................................................................................................................37  Factorial  Validity  and  Reliability................................................................................................................................................38  Summary  and  Recommendations .............................................................................................................................................41  

STUDENT  SURVEY  (GRADES  5-­‐8) ...........................................................................................................................................................43  

Respondent  Characteristics .......................................................................................................................................................43  Response  Pattern  Summaries....................................................................................................................................................43  Factorial  Validity  and  Reliability................................................................................................................................................44  Summary  and  Recommendations .............................................................................................................................................47  

STUDENT  SURVEY  (GRADES  9-­‐12) .........................................................................................................................................................49  

Page 3: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  3      

Respondent  Characteristics .......................................................................................................................................................49  Response  Pattern  Summaries....................................................................................................................................................49  Factorial  Validity  and  Reliability................................................................................................................................................50  Summary  and  Recommendations .............................................................................................................................................52  

APPENDIX................................................................................................................................................................................ 54  

APPENDIX  A:  TABLES  FROM  THE  ITEM-­‐LEVEL  ANALYSES  OF  THE  CATHOLIC  SCHOOL  –  PROGRAM  EFFECTIVENESS  SURVEYS.......................................54  APPENDIX  B:  TABLES  FROM  THE  ITEM-­‐LEVEL  ANALYSES  OF  THE  CATHOLIC  SCHOOL  –  DEFINING  CHARACTERISTICS  SURVEYS ....................................75  

 

Page 4: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  4      

 

Executive  Summary  Arising  from  a  widely  recognized  need  by  the  community  of  Catholic  educators  for  “a  common  framework  of  universal  characteristics  of  Catholic  identity  and  agreed  upon  criteria  for  Catholic  school  excellence,”  the  National  Standards  and  Benchmarks  for  Effective  Catholic  Elementary  and  Secondary  Schools  was  published  in  2012.1  This  publication  delineates  a  set  of  Defining  Characteristics  of  Catholic  schools:    (a)  Centered  in  the  Person  of  Jesus  Christ,  (b)  Contributing  to  the  Evangelizing  Mission  of  the  Church,  (c)  Distinguished  by  Excellence,  (d)  Committed  to  Educate  the  Whole  Child,  (e)  Steeped  in  a  Catholic  Worldview,  (f)  Sustained  by  Gospel  Witness,  (g)  Shaped  by  Communion  and  Community,  (h)  Accessible  to  All  Students,  and  (j)  Established  by  the  Expressed  Authority  of  the  Bishop.  These  Defining  Characteristics  provide  the  foundation  for  a  set  of  four  Standards  and  Benchmarks  for  Catholic  schools.  The  Standards  articulate  policies,  programs,  structures,  and  processes  in  four  domains  –  Mission  and  Catholic  Identity,  Governance  and  Leadership,  Academic  Excellence,  and  Operational  Vitality  –  that  “should  be  present  in  mission-­‐driven,  program  effective,  well-­‐managed,  and  responsibly  governed  Catholic  schools….”  Subsequently,  the  task  force  that  authored  the  National  Standards  and  Benchmarks  for  Effective  Catholic  Elementary  and  Secondary  Schools  document  developed  a  set  of  survey  instruments  to  measures  schools’  alignment  with  the  Defining  Characteristics  and  the  Standards  for  effective  Catholic  schools  through  the  perceptions  of  school-­‐affiliated  adults  and  students  (5th  –  12th  grades)  and,  in  collaboration  with  AdvancED,  designed  a  psychometric  study  to  gauge  the  validity  and  reliability  of  the  survey  instruments.  The  focus  of  this  technical  report  is  on  that  psychometric  study  and  analyses  of  the  newly  developed  survey  instruments.      Participating  were  nearly  8,000  adults  and  students  from  55  Catholic  primary  and  secondary  schools.  Multilevel,  ordinal  factor  analyses  were  conducted  to  examine  the  validity  and  reliability  of  the  survey  instruments.  These  analyses  sought  to  elucidate  the  number  and  pattern  of  factors  being  measured  by  the  surveys,  including  the  congruence  of  the  factor  structure  with  the  guiding  theoretical  framework;  the  validity  and  reliability  of  its  items;  and  the  reliability  and  distribution  of  scale  and  subscale  composite  scores.  Based  on  the  results  of  this  study,  each  survey  reliably  captured  a  single  factor  or  dimension  reflecting  the  Defining  Characteristics  or  Standards  for  effective  Catholic  schools.  This  was  perhaps  contrary  to  expectation  for  the  Catholic  School  Program  Effectiveness  Surveys  (Standards),  for  which  four  distinct  factors  relating  to  the  four  domains  articulated  in  the  Standards  (viz.,  Mission  and  Catholic  Identity,  Governance  and  Leadership,  Academic  Excellence,  and  Operational  Vitality)  were  expected.  Most  items  were  supported  by  the  analyses  as  valid  and  reliable  indicators  of  their  respective  factors,  and  reliability  estimates  of  a  composite  (total)  score  and  most  subscale  scores  computed  by  averaging  over  respondents  

                                                                                                               

1  Center  for  Catholic  School  Effectiveness,  School  of  Education,  Loyola  University  Chicago,  in  partnership  with  the  Barbara  and  Patrick  Roche  Center  for  Catholic  Education,  School  of  Education,  Boston  College.  (2012).  National  Standards  and  Benchmarks  for  Effective  Catholic  Elementary  and  Secondary  Schools.  http://www.catholicschoolstandards.org/  

Page 5: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  5      

and  items  within  the  same  school  indicated  exceptionally  high  reliability.2  The  one  exception  was  the  Operational  Vitality  subscale  on  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  –  8th  grades);  the  reliability  for  the  scale  was  marginal.  Although  nearly  all  items  were  found  to  be  reliable  indicators  of  their  respective  factors,  all  surveys  had  at  least  one  item  with  missing  data  rates,  due  to  either  Don’t  Know  responses  or  being  skipped,  exceeding  10%  and  some  approaching  or  exceeding  40%.  These  items  can  be  problematic  as  high  rates  of  missing  data  negatively  affect  the  internal  validity  and  generalizability  of  the  survey  measurements.  The  detailed  technical  report  that  follows  elaborates  on  the  sampling  and  statistical  methodology  and  the  psychometric  analyses  of  these  surveys.  The  results  from  the  psychometric  analyses  are  described  in  multiple  sections,  each  with  parallel,  self-­‐contained  subsections,  pertaining  to  each  of  the  six  surveys.  These  sections  are  concluded  with  a  summary  and  recommendations  for  scale  interpretation  and  possible  revisions  to  the  surveys.  

                                                                                                               

2  Due  to  a  limited  sample  of  high  schools,  the  9th  –  12th  grade  surveys  were  examined  only  at  the  respondent  (student)  level,  which  is  not  the  intended  level  of  measurement  for  these  surveys.  As  such,  inferences  about  the  validity  and  reliability  of  the  9th  –  12th  grade  surveys  drawn  from  this  study  are  limited.  

Page 6: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  6      

 

Sampling  and  Methodology  Sampling  Overview  

 During  January  through  March  of  2012,  a  sample  of  200  Catholic  schools  was  selected  to  receive  an  invitation  to  participate  in  the  pilot  study  by  administering  a  battery  of  surveys  to  the  students  (5th  –  12th  grades)  and/or  adults  (school-­‐based  and  non-­‐school-­‐based)  affiliated  with  their  schools.  Of  the  200  schools  invited,  55  schools  agreed  to  participate  and  had  at  least  one  respondent  for  at  least  one  of  the  surveys.  Eligible  respondents  (all  adults  and  students  between  5th  and  12th  grades  affiliated  with  the  participating  schools)  were  invited  to  complete  the  Catholic  School  Program  Effectiveness  Survey  and/or  Catholic  School  Defining  Characteristics  Survey.  Separate  versions  of  each  survey  were  administered  to  adults,  5th  –  8th  grade  students,  and  9th  –  12th  grade  students.  Characteristics  of  the  samples  differed  across  the  six  surveys,  and  so  individual  respondent  characteristics  are  provided  in  the  survey-­‐specific  sections  found  later  in  this  report.  Survey  administration  was  conducted  through  a  web-­‐interface  using  Survey  Monkey,  and  participation  was  anonymous.    

Methodology  

Analyses  began  with  a  simple  examination  of  item  statistics  at  the  respondent  (adult/student)  and  school  level.3  These  analyses  were  focused  on  describing  the  distribution  of  responses  for  each  item,  with  an  eye  toward  the  variability  in  responses  across  schools  and  across  items.  Inter-­‐item  variability  is  desirable  for  reliable  measurement  of  the  Standards  for  Program  Effectiveness  and  Defining  Characteristics  of  Catholic  Schools  across  the  range  of  scores  on  the  respective  factors.  If  most  or  all  items  are  endorsed  with  high  agreement  responses,  the  items  collectively  might  discriminate  reliably  amongst  schools  scoring  high  on  these  factors,  but  not  reliably  for  schools  intermediate  or  relatively  weak  on  these  factors.    To  evaluate  the  construct  validity  of  these  scales,  factor  analytic  techniques  were  employed  to  evaluate  their  factorial  validity  and  reliability.  Factor  analysis  seeks  to  ascertain  the  underlying  (i.e.,  latent,  unobserved)  structure  of  the  measurement  instrument  (e.g.,  survey)  and  is  an  important  prerequisite  before  other  components  of  construct  validity  or  reliability  estimation  are  conducted.    Examples  of  the  questions  that  factor  analysis  can  address  include:  

• How  well  does  the  underlying  factor  dimensionality  and  structure  align  with  the  theory  that  guided  the  development  of  the  surveys?    

Are  there  four  dimensions  corresponding  to  the  four  Standards  for  Program  Effectiveness  of  Catholic  Schools?  Is  there  a  single  dimension  capturing  the  Defining  Characteristics  for  Catholic  Schools?  If  not,  how  many  dimensions  are  measured  by  the  surveys?  

Which  items  reliably  measure  which  dimension(s)?  • Are  the  identified  factors  reliably  measured  by  the  indicators  (i.e.,  items)?  Are  items  all  valid  indicators  

of  the  underlying  construct(s)?  

                                                                                                               

3  Due  to  a  limited  sample  of  schools,  school-­‐level  analysis  was  not  possible  for  the  9th-­‐12th  grade  surveys.  

Page 7: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  7      

Which  items,  if  any,  need  to  be  discarded  or  revised  due  to  poor  validity  or  reliability  or  due  to  measuring  more  than  one  dimension  (item  complexity)?    

Do  some  items  convey  redundant  information  about  the  underlying  construct  and  thus  can  be  discarded  without  loss  of  information  (reliability)  to  ease  response  burden?  

Adult  and  students  responses  to  the  surveys  were  subjected  to  a  multilevel,  confirmatory  factor  analysis  (CFA)  to  ascertain  the  degree  of  statistical  fit  between  the  data  and  models  specified  in  accord  with  the  theoretical  intent  of  the  survey's  design.4  As  it  was  considered  a  possibility  that  neither  the  four-­‐factor  model  nor  the  one-­‐factor  model  would  accurately  depict  the  factor  structure  of  the  surveys,  an  exploratory  factor  analysis  (EFA)  was  also  performed.  EFA  allows  the  scale  developer  to  ascertain  the  degree  of  statistical  fit  between  the  observed  data  and  a  model  with  k  factors,  where  the  range  of  k  examined  depended  on  the  eigenvalues  from  the  reduced  correlation  matrix.  Interpretability  of  the  solution  and  statistical  tests  and  indices  of  model  fit  were  used  to  settle  on  a  particular  k  factor  solution.  Exploratory  factor  analytical  methods  impose  minimal  a  priori  constraints  (hypotheses/predictions  on  the  model),  beyond  those  required  to  statistical  identify  the  estimated  parameters  of  the  model,  and  are  aimed  at  building  a  model  of  the  underlying  factor  structure  in  the  absence  of  information  about  the  structure.  This  analysis  employed  the  Geomin  rotation,  which  allows  factors  to  correlate  in  models  where  k  >  1.  All  factor  analyses  were  conducted  using  the  Mplus  statistical  software  program  (v.  6.12)5  and  employed  a  mean-­‐  and  variance-­‐adjusted  weighted  least  squares  estimator  with  numerical  integration.    Model  fit  tests  and  indices  used  consist  of  the  chi-­‐square  test  of  exact  fit,  the  Comparative  Fit  Index  (CFI),  the  root  mean  square  error  of  approximation  (RMSEA),  and  standardized  root  mean  residual  (SRMR  –  school-­‐level  model).  Although  statisticians  continue  to  debate  the  appropriate  focus  and  thresholds  on  these  tests  and  fit  indices,  most  consider  a  statistically  non-­‐significant  chi-­‐square  test  of  exact  fit  (reflected  by  probability  values  greater  than  .05),  RMSEA  values  less  than  .06,  CFI  values  greater  than  .96,  and  SRMR  values  below  .06  to  reflect  good  or  adequate  model  fit  to  the  data.  In  other  words,  a  model  with  k  factors  that  meets  these  criteria  is  said  to  be  consistent  with  the  data  and  therefore  accepted  for  further  consideration.  Conversely,  a  model  that  fails  all  criteria  is  said  to  be  rejected  by  the  data.  In  practice,  models  often  meet  only  some  of  the  criteria,  while  being  near  but  just  outside  the  thresholds  for  the  acceptable  range  on  other  criteria.  In  these  cases,  the  theoretical  interpretability  of  the  results  dominates.  When  two  or  more  models  exhibit  similar  fit,  the  model  that  is  most  interpretable  and  parsimonious  is  usually  retained.  For  models  that  fit  the  data  well,  item-­‐level  statistics  are  examined  to  evaluate  the  validity  and  reliability  of  individual  items.    These  analyses  account  for  two  important  characteristics  of  the  data:  (1)  response  data  are  collected  using  a  Likert  scale  and  thus  may  not  possess  interval  scale  properties,  and  (2)  responses  to  the  surveys  are  nested  within  schools  and  thus  are  not  independently  distributed  (i.e.,  since  respondents  affiliated  to  the  same  school  are  reporting  on  the  same  school,  they  are  more  correlated  than  with  responses  from  respondents  affiliated  with  different  schools),  an  assumption  of  standard  factor  analytic  methods.  With  respect  to  (1),  common  analytical  methods  (Pearson-­‐product  moment  correlation  analysis,  linear  factor  analysis)  often  employed  with  survey  data  assume  that  the  scale  of  measurement  for  the  data  are  interval-­‐

                                                                                                               

4  Due  to  a  limited  sample  of  schools,  a  single-­‐level  analysis  at  the  respondent-­‐level  was  conducted  for  the  9th-­‐12th  grade  surveys.  5  Muthén  &  Muthén,  2011,  http://www.statmodel.com  

Page 8: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  8      

level  (differences  in  adjacent  response  options  reflect  equal  discriminations  on  the  underlying  agreement  scale  used  by  the  respondent).  As  such,  inappropriate  use  of  these  linear  methods  with  binary  or  ordinal  data  can  lead  to  statistical  artifacts  and  biased  results,  particularly  when  the  number  of  response  options  is  fewer  than  seven  and/or  the  response  distribution  is  highly  skewed  (e.g.,  strong  floor  or  ceiling  effects).  To  address  this  limitation  of  conventional,  linear  factor  analytic  approaches,  categorical  confirmatory  factor  analysis  methods6  were  employed.  Instead  of  modeling  the  variances  and  covariances  (and  means)  of  the  response  data  as  a  linear  function  of  the  latent  constructs  given  the  model  parameters:   ,  

one  parameterization  for  ordinal  factor  analysis  using  a  probit  link  function  models  the  item-­‐response  probability  distribution  as  a  non-­‐linear  function  of  the  latent  constructs  (given  the  model  parameters)  via  a  latent  response  distribution.  In  this  model,  latent  response  variates,  y*,  are  conceptualized  to  reflect  a  latent,  interval-­‐scaled  variable  underlying  each  item,  y,  such  that   ,  where  c  

represents  the  categories  (e.g.,  never,  seldom,  …)  of  y  and    are  the  estimated  threshold  parameters  for  the  probit  function  linking  y*  with  y.  Based  on  the  assumption  of  bivariate  normality  of  y*  (and  other  regularity  assumptions)  the  estimated  variance-­‐covariance  matrix  of  y*  is  modeled  as  a  function  of  the  

underlying  latent  factors  given  the  estimated  and  constrained  model  parameters:   .  

With  regards  to  (2),  the  ordinal  factor  analytic  models  described  above  were  extended  to  account  for  the  nested  (multilevel  or  hierarchical)  structure  of  the  data  (i.e.,  respondents  are  nested  within  schools)  in  line  with  the  approaches  described  by  Grilli  and  Rampichini7    and  Asparouhov  and  Muthén.8      Once  an  acceptable  factor  model  was  established,  an  estimate  of  the  reliability  (internal  consistency)  for  the  scales  and  subscales  was  calculated  using  an  approach  described  by  Raykov.9  Raykov’s  internal  consistency  reliability  coefficient  captures  the  precision  of  a  scale  score  (e.g.,  computed  by  summing  or  averaging  all  items  within  the  scale)  and  is  similar  to  Cronbach’s  alpha  coefficient.  Both  coefficients  typically  range  between  0  and  1  (with  coefficients  approaching  1  being  indicative  of  more  precise  measurement  and  values  greater  than  0.8  being  preferred).  Values  are  interpreted  as  the  proportion  of  scale  variance  that  is  “true  score”  variance  deriving  from  the  underlying  common  factor  and  which  is  not  due  to  random  measurement  error  or  item  specific  variance.  However,  Cronbach’s  alpha  coefficient  is  unbiased  only  under  a  fairly  stringent  assumption  about  the  relationship  between  the  underlying  factor  and  the  observed  item  responses  (viz.  essential  tau  equivalence).  When  this  assumption  is  not  met,  as  is  often  the  case,  Cronbach’s  alpha  is  a  lower-­‐bound  estimate  of  reliability,  though  another  issue  can  lead  to  upward  bias.  

                                                                                                               

6  Flora,  D.  B.,  &  Curran,  P.  J.  (2004).  An  Empirical  Evaluation  of  Alternative  Methods  of  Estimation  for  Confirmatory  Factor  Analysis  with  Ordinal  Data.  Psychological  Methods,  9,  466-­‐491.  doi:10.1037/1082-­‐989X.9.4.466  

Wirth,  R.  J.,  &  Edwards,  M.  C.  (2007).  Item  factor  analysis:  Current  approaches  and  future  directions.  Psychological  Methods,  12,  58-­‐79.  doi:10.1037/1082-­‐989X.12.1.58  7  Grilli,  L.,  &  Rampichini,  C.  (2007).  Multilevel  factor  models  for  ordinal  variables.  Structural  Equation  Modeling,  14(1),  1-­‐25.  doi:10.1207/s15328007sem1401_1  8  Asparouhov,  T.,  &  Muthén,  B.  (2007).  Computationally  efficient  estimation  of  multilevel  high-­‐dimensional  latent  variable  models.  Proceedings  of  the  2007  Joint  Statistical  Meetings,  Section  on  Statistics  in  Epidemiology  (pp.  2531–  2535).  Alexandria,  VA:  American  Statistical  Association. 9  Raykov,  T.  (1997).  Scale  reliability,  Cronbach's  Coefficient  Alpha,  and  violations  of  essential  tau-­‐equivalence  with  fixed  congeneric  components.  Multivariate  Behavioral  Research,  32(4),  329-­‐353.  doi:10.1207/s15327906mbr3204_2  

Page 9: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  9      

Raykov’s  reliability  coefficient  (rho)  is  valid  under  a  weaker  assumption  of  congeneric  structure  and  can  account,  if  properly  modeled,  for  a  common  source  of  upward  bias.  

Page 10: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  10      

 

Program  Effectiveness  Surveys  Description  

The  Program  Effectiveness  battery  of  surveys  is  designed  to  gauge  the  extent  to  which  Catholic  Schools  adhere  to  standards  developed  by  the  Task  Force  for  National  Standards  and  Benchmarks  for  Effective  Catholic  Elementary  and  Secondary  Schools.10  Separate  self-­‐report  surveys  were  designed  to  capture  the  perceptions  of  adults,  middle  school  (5th  –  8th  grade)  students,  and  high  school  (9th  –  12th  grade)  students  of  the  “policies,  programs,  structures,  and  processes  that  should  be  present  in  mission-­‐driven,  program  effective,  well-­‐managed,  and  responsibly  governed  Catholic  schools,”  according  to  the  National  Standards  and  Benchmarks  for  Effective  Catholic  Elementary  and  Secondary  Schools.  Each  survey  has  three  or  more  items  pertaining  to  each  of  the  four  domains  articulated  in  the  Standards:  Mission  and  Catholic  Identity,  Governance  and  Leadership,  Academic  Excellence,  and  Operational  Vitality.  The  Catholic  School  Program  Effectiveness  Survey  of  Adults  consists  of  42  items,  with  Mission  and  Catholic  Identity,  Governance  and  Leadership,  Academic  Excellence,  and  Operational  Vitality  subscales  consisting  of  13,  7,  14,  and  8  items,  respectively.  The  Catholic  School  Program  Effectiveness  Survey  of  5th  –  8th  grade  students  consists  of  27  items,  with  Mission  and  Catholic  Identity,  Governance  and  Leadership,  Academic  Excellence,  and  Operational  Vitality  subscales  consisting  of  9,  3,  12,  and  3  items,  respectively.  The  Catholic  School  Program  Effectiveness  Survey  of  9th  –  12th  grade  students  consists  of  35  items,  with  Mission  and  Catholic  Identity,  Governance  and  Leadership,  Academic  Excellence,  and  Operational  Vitality  subscales  consisting  of  13,  4,  14,  and  4  items,  respectively.  Survey  respondents  self-­‐reported  their  perceptions  on  a  5-­‐point  Likert  response  scale  ranging  from  Strongly  Disagree  (1)  to  Strongly  Agree  (5).  Specific  items  and  their  subscale  grouping  can  be  found  in  Appendix  A:  Tables  from  the  Item-­‐level  Analyses  of  the  Catholic  School  –  Program  Effectiveness  Surveys.  

Adult  Survey  

Respondent  Characteristics  

A  total  of  3,526  respondents  from  55  schools  (average  of  64  respondents  per  school)  completed  the  Catholic  School  Program  Effectiveness  Survey  of  Adults  for  this  psychometric  study.  The  majority  of  respondents  were  non-­‐school  based  (75%).  Of  non-­‐school  based  respondents,  94.6%  were  parents,  2.9%  were  community  members,  and  2.5%  self-­‐identified  as  “other”.  Among  the  parent  respondents,  40.2%  reported  their  oldest  child’s  grade  as  being  4th  grade  or  below,  38%  in  5th  –  8th  grades,  and  12%  in  9th  –  12th  grades.  Of  school-­‐based  respondents  (25%  of  the  total  sample),  71.8%  were  teachers,  19.1%  were  staff,  and  9.1%  were  administrators.  Most  respondents  were  female  (83%),  and  the  racial  composition  of  the  sample  was:  White  (92%),  African  American  or  Black  (4.5%),  multi-­‐racial  (1.7%),  Asian  (1.2%),  American  Indian  or  

                                                                                                               

10  Center  for  Catholic  School  Effectiveness,  School  of  Education,  Loyola  University  Chicago,  in  partnership  with  the  Barbara  and  Patrick  Roche  Center  for  Catholic  Education,  School  of  Education,  Boston  College.  (2012).  National  Standards  and  Benchmarks  for  Effective  Catholic  Elementary  and  Secondary  Schools.  http://www.catholicschoolstandards.org/  

Page 11: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  11      

Alaska  Native  (.3%),  or  Native  Hawaiian  or  Pacific  Islander  (.1%).  The  majority  of  respondents  identified  as  non-­‐Hispanic  (95.8%).    Most  respondents  expressed  their  religious  affiliation  as  Catholic  (85.7%).  In  general,  respondents  reported  multiple  years  of  affiliation  with  the  school:  30.7%  reported  1-­‐4  years,  31.5%  reported  5-­‐10  years,  and  27.8%  reported  more  than  10  years.  Only  10%  reported  less  than  1  year  of  affiliation  with  the  school.  

Response  Pattern  Summaries  

Response  frequencies  and  intraclass  correlation  coefficients  (ICCs)  for  each  of  the  42  items  are  reported  in  Table  A.1  in  Appendix  A:  Tables  from  the  Item-­‐level  Analyses  of  the  Catholic  School  –  Program  Effectiveness  SurveysAppendix  .  On  26  of  the  items,  at  least  75%  of  adults  endorsed  the  Strongly  Agree  or  adjacent  response  options,  and  all  items  were  endorsed  (Strongly  Agree  or  adjacent  response)  by  at  least  50%  of  respondents.  Though  response  frequencies  are  clustered  towards  endorsement  of  options  indicating  greater  agreement,  there  was  some  degree  of  variability  of  frequency  patterns  across  items.  Item  percentages  for  skipped  or  Don't  Know  responses  ranged  between  9.7%  and  43.1%  (Mean  =  21.4%).  Thirteen  items  were  skipped  or  had  a  Don't  Know  response  from  25%  or  more  of  respondents.  All  Operational  Vitality  items  exhibited  rates  exceeding  28%.  The  most  extreme  instance  was  item  35  (Our  school’s  financial  plan  is  the  result  of  a  collaborative  process  including  expert  advisors).  Of  note  is  that  the  rates  of  using  the  Don’t  Know  option  were  much  higher  for  these  items  than  most  preceding  items,  which  may  indicate  that  respondents  did  not  feel  they  had  sufficient  knowledge  of  these  items  to  provide  their  ratings.  However  explained,  these  items  should  be  considered  for  rewording,  replacement  or  deletion  from  the  scale,  along  with  other  items  exhibiting  the  highest  skip  or  Don't  Know  rates,  given  the  problems  missing  data  presents  for  estimation  of  subscale  and  scale  scores.11    ICCs  quantify  the  proportion  of  variability  in  responses  that  is  attributable  to  variability  between  scores.  In  other  terms,  an  ICC  reflects  the  degree  of  non-­‐independence  amongst  responses  from  adults  affiliated  with  the  same  school,  with  0  =  independence  (i.e.,  no  systematic  variation  across  schools)  and  1  =  complete  dependence  (i.e.,  all  variation  in  responses  is  due  to  differences  across  schools).  As  this  survey  is  intended  to  measure  school-­‐level  adherence  to  Standards,  ICCs  greater  than  zero  are  to  be  expected  and  desired.  Additionally,  ICCs  greater  than  .01  support  the  need  for  statistical  methods  that  account  for  the  observed  non-­‐independence.  The  ICCs  for  most  items  were  moderate  in  magnitude,  reflecting  similarity  in  responses  from  adults  affiliated  the  same  school  and  justifying  the  need  for  statistical  methods  that  can  account  for  non-­‐independence  amongst  the  observations.  ICCs  across  all  items  ranged  from  .03  to  .11  (Mean  =  .08),  indicating  that  between  3%  and  11%  of  variation  in  item  responses  was  attributable  to  respondents  being  affiliated  with  different  schools.    The  Catholic  School  Program  Effectiveness  Survey  of  Adults  is  intended  to  provide  data  on  Catholic  schools’  adherence  to  the  Standards.  Given  this  intent,  the  focus  of  these  analyses  is  on  school-­‐level  aggregations  of  the  responses.  Table  A.2  in  Appendix  A:  Tables  from  the  Item-­‐level  Analyses  of  the    provides  statistics  to  describe  the  distributions  of  these  school-­‐level  aggregates,  where  adult  responses  for  each  item  are  

                                                                                                               

11  Although  most  analyses  described  in  the  Factorial  Validity  and  Reliability  use  statistical  procedures  that  can  incorporate  respondents  that  are  missing  data  on  one  or  more  items,  respondents  missing  data  on  all  items  are  necessarily  omitted  from  these  analyses.  Depending  on  the  particular  analysis,  between  286  and  694  respondents  were  omitted  from  subsequent  analysis  due  to  missing  data  on  all  items.  

Page 12: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  12      

averaged  with  other  adults  from  their  school.  The  theoretical  range  for  these  aggregated  item  averages  is  1  (Strongly  Disagree)  -­‐  5  (Strongly  Agree),  with  higher  scores  indicating  higher  within-­‐school  average  agreement  for  the  particular  item.  On  average,  schools  were  rated  highly  on  all  of  the  items  (mean  average  rating  across  items  =  4.08).  Along  with  average,  minimum,  and  maximum  ratings,  standard  deviations  and  quartiles  are  also  provided  for  each  item.  The  median  (also  known  as  the  50th  percentile  or  second  quartile)  reflects  the  item  score  at  which  50%  of  schools  fall  at  or  below.  For  all  items,  50%  or  more  of  the  schools  scored  at  least  a  3.59.  Similarly  to  the  respondent-­‐level  item  statistics,  scores  were  clustered  toward  the  upper  end  of  the  distribution.  

Factorial  Validity  and  Reliability  

Multilevel,  ordinal  confirmatory  factor  analyses  (CFA)  began  with  the  separate  examination  of  Catholic  School  Program  Effectiveness  Survey  of  Adults  subscales  corresponding  to  each  of  the  four  Standards  for  Catholic  School  Program  Effectiveness.  These  CFA  models,  consisting  of  one-­‐factor  with  seven  to  fourteen  indicators  each,  were  fit  to  the  data.  Then,  these  models  were  combined  into  a  single,  four-­‐factor  CFA  model  of  the  Catholic  School  Program  Effectiveness  Survey  of  Adults.  CFA  concluded  with  estimation  of  a  model  where  all  items  were  specified  as  measures  of  a  single  factor  (one-­‐factor  CFA  model).12  The  model  fit  tests  and  indices  for  these  analyses  are  reported  in  Table  1Error!  Not  a  valid  bookmark  self-­‐reference..  These  tests  evaluate  how  well  each  model  fits  the  data.  The  aim  is  to  identify  a  model  that  fits  the  data  well  (within  sampling  error)  or  approximately  well  and  is  parsimonious.  If  a  model  fits  the  data  poorly,  the  model  would  be  rejected  from  further  consideration.    

Table  1.  Model  Fit  Tests  and  Indices  for  Full-­Scale  and  Subscale  CFA  Models  of  the  Catholic  School  Program  Effectiveness  Survey  of  Adults  

 Chi-­‐Square   df   p-­‐value   RMSEA   CFI   SRMR  

Mission  and  Catholic  Identity  

84.11   65   0.06   0.010   1.00   0.078  

Governance  and  Leadership  

30.77   14   0.006   0.020   1.00   0.039  

Academic  Excellence   93.31   77   0.10   0.008   1.00   0.056  

Operational  Vitality   28.09   20   0.10   0.012   1.00   0.040  

1-­‐Factor  Model   889.37   819   0.04   0.005   1.00   0.080  

4-­‐Factor  Model   -­‐-­‐   -­‐-­‐   -­‐-­‐   -­‐-­‐   -­‐-­‐   -­‐-­‐  

Note:  df  =  degrees  of  freedom;  RMSEA  =  Root  Mean  Square  Error  of  Approximation;  CFI  =  Comparative  Fit  Index;  SRMR  =  Standardized  Root  Mean  Residual.    

                                                                                                               

12  In  these  analyses,  model  parameters  are  being  simultaneously  estimated  at  two-­‐levels,  respondent-­‐level  and  the  school-­‐level.,  with  an  unstructured  model  specified  for  the  within-­‐level  (i.e.,  no  over-­‐identifying  constraints  included).  However,  the  focus  of  these  analyses  is  on  discovering  the  factors  that  pertain  to  schools.    Thus,  the  focus  of  this  report  is  exclusively  on  results  pertaining  to  the  school-­‐level  models.  

Page 13: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  13      

According  to  the  conventional  guidelines  described  earlier  in  the  methodology  section,  all  subscale  models  except  for  the  Governance  and  Leadership  model  met  exact  fit  standards.  Although  the  Governance  and  Leadership  model  exhibited  some  degree  of  statistical  misfit,  this  model  fit  the  data  approximately  well  according  to  the  RMSEA,  CFI,  and  SRMR  fit  indices.  The  four-­‐factor  CFA  model,  which  combined  the  subscale  models  into  a  single  analysis,  did  not  converge  to  a  solution  when  the  specified  number  of  iterations.  This  might  be  due  to  high  collinearity  amongst  the  factors  as  noted  below.  Therefore,  a  one-­‐factor  CFA  model  was  examined.  As  reflected  by  the  model  fit  indices,  the  one-­‐factor  model  fit  the  data  approximately  well.  Still,  it  is  essential  that  the  one-­‐factor  model  solution  be  considered  meaningful  and  interpretable  where  the  meaningfulness  or  interpretability  of  a  solution  is  determined  by  considering  the  strength  and  pattern  of  relationships  between  the  items  and  underlying  (latent)  factor.  The  relationships  between  factors  and  indicators  are  typically  depicted  in  a  factor  loading  matrix.  Standardized  factor  loadings  and  associated  standard  errors  for  the  one-­‐factor  CFA  model  and  for  each  one-­‐factor  subscale  model  are  provided  in  Table  in  Appendix  A:  Tables  from  the  Item-­‐level  Analyses  of  the  .  The  standardized  factor  loadings  were  uniformly  very  high,  with  the  vast  majority  greater  than  0.8  and  none  below  0.58.  These  numbers  indicate  that  each  item  is  a  salient  and  highly  reliable  measure  of  the  factor.  Moreover,  the  standardized  factor  loadings  are  estimated  with  a  moderate  degree  of  precision,  as  reflected  by  relatively  small  standard  errors.      Though  the  CFA  analyses  and  higher  inter-­‐subscale  correlations  suggest  a  one-­‐factor  model,  it  is  possible  that  another  k-­‐factor  model  not  examined  generated  the  data.  Therefore,  an  exploratory  factor  analysis  (EFA)  was  performed  prior  to  settling  on  the  one-­‐factor  model  of  the  Catholic  School  Program  Effectiveness  Survey  of  Adults.  An  important  and  initial  task  in  conducting  exploratory  factor  analysis  is  to  determine  the  number  of  factors  or  dimensions  that  are  being  measured  by  the  survey  instrument.  This  determination  is  guided  by  statistical  tests  and  indices,  evaluation  of  eigenvalues,  and  meaningfulness  and  interpretability  of  the  solution.  As  with  the  confirmatory  factor  analyses,  there  are  two-­‐levels  under  consideration,  respondent-­‐level  and  the  school-­‐level,  and  the  number  of  factors  would  typically  be  determined  at  each  level.  However,  the  focus  of  this  report  is  on  the  school  as  the  unit  of  analysis.        Eigenvalues  quantify  the  variance  in  the  item  responses  that  is  explained  by  the  factors.  Factors  that  account  for  more  variation  are  considered  potentially  more  important  or  meaningful  than  factors  that  account  for  less  variation.  Eigenvalues  from  the  school-­‐level  factor  analysis  of  the  survey  data  are  reported  in  Table  2.  An  often  cited  rule  of  thumb  is  the  Kaiser  rule,  which  states  that  factors  with  eigenvalues  greater  than  1.0  should  be  extracted.  According  to  the  Kaiser  rule,  a  model  with  four  factors  should  be  examined.  This  rule,  however,  has  been  criticized  as  leading  to  extraction  of  too  many  factors  and  thus  typically  should  be  considered  an  upper  bound  estimate  of  the  number  of  factors.      

Table  2.  Eigenvalues  for  the  Exploratory  Factor  Analysis  of  the  Catholic  School  Program  Effectiveness  Survey  of  Adults  

    Factor   1   2   3   4   5   6   7   8   9   10   ...  

Eigenvalues       31.90   2.22   2.07   1.43   0.95   0.87   0.65   0.58   0.47   0.36    

 Another  factor  enumeration  approach  used  is  examination  of  the  Scree  plot,  depicted  in  Figure  1.  With  this  approach,  one  typically  seeks  the  point  where  there  is  a  pronounced  bend  (elbow)  in  the  curve.  Factors  before  the  bend  are  given  further  consideration.  According  to  the  scree  plot,  a  one-­‐factor  or  perhaps  a  two-­‐factor  model  should  be  considered  for  extraction.  Although  more  sophisticated  methods  for  factor  

Page 14: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  14      

enumeration  exists  (e.g.,  parallel  analysis),  these  were  computationally  infeasible  or  inaccessible  for  the  present  analysis,  which  involve  a  multilevel  structure  and  ordinal  item  responses  distributions  that  make  it  difficult  to  employ  more  advanced  methods.  Thus,  the  one-­‐factor  and  two-­‐factor  models  were  given  closer  consideration.  

Figure  1.  Scree  Plot  of  Eigenvalues  (School-­Level)  for  the  Catholic  School  Program  Effectiveness  Survey  of  Adults  

   In  conjunction  with  examination  of  the  eigenvalues  and  scree  plot,  statistical  tests  and  indices  of  model  fit  are  often  consulted.  All  models  fit  the  data  well  (see  Table  3).  With  most  indices  being  within  desired  ranges  for  all  models  under  consideration,  the  philosophical  principle  known  as  Occam's  razor  dictates  that  one  would  choose  the  most  parsimonious  model,  which  often  is  the  model  with  the  fewest  factors.  Consistent  with  the  CFA,  model  fit  indices,  in  conjunction  with  the  eigenvalues,  scree  plot  and  factor  pattern  matrix,  support  a  model  that  posits  a  single  factor  for  explaining  and  summarizing  the  Catholic  School  Program  Effectiveness  Survey  of  Adults  item  responses  when  aggregated  to  the  school-­‐level.  This  conclusion  was  corroborated  by  the  less  interpretable  solutions  for  the  EFA  models  with  two  or  more  factors  (e.g.,  factor  loadings  >  1,  most  items  load  on  a  general  factor,  with  some  Academic  Excellence  and  other  items  cross-­‐loading  onto  a  second  factor).  

Table  3.  Model  Fit  Tests  and  Indices  for  EFA  Models  of  the  Catholic  School  Program  Effectiveness  Survey  of  Adults  

 1-­‐Factor  Model  

2-­‐Factor  Model  

3-­‐Factor  Model  

4-­‐Factor  Model  

Chi-­‐Square   889.37   812.09   715.60   652.03  

df   819   778   738   699  

0  

5  

10  

15  

20  

25  

30  

35  

1   2   3   4   5   6   7   8   9   10  

Eigenvalue

 

Factor  

Page 15: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  15      

p-­‐value   0.04   0.19   0.71   0.90  

RMSEA   0.005   0.004   <0.001   <0.001  

CFI   1.00   1.00   1.00   1.00  

SRMR   0.080   0.065   0.047   0.039  

Summary  statistics  for  the  distribution  of  composite  scores  formed  by  taking  the  average  of  school-­‐level  item  averages  for  the  entire  Catholic  School  Program  Effectiveness  Survey  of  Adults  and  each  subscale  can  be  found  in  Table  4.  The  theoretical  range  for  the  composite  and  subscale  scores  is  1  -­‐  5,  with  5  reflecting  greater  adherence  to  the  Standards  as  reported  by  the  respondents.  On  average,  schools  tended  to  score  highly  on  the  scale  and  each  subscales,  with  75%  or  more  schools  scoring  greater  than  3.69  on  all  scales.    

Table  4.  Descriptive  Statistics  for  a  Total  Scale  Composite  Score  and  Subscale  Scores  (School-­level)  Derived  from  the  Catholic  School  Program  Effectiveness  Survey  of  Adults  

 

Mean Standard Deviation Minimum Maximum 25th

Percentile Median 75th Percentile Reliability SEM

Mission  and  Catholic  Identity  

4.15   0.26   3.39   4.65   4.01   4.15   4.32   0.98   0.04  

Governance  and  Leadership  

4.22   0.33   2.94   4.74   4.09   4.23   4.42   0.99   0.03  

Academic  Excellence  

4.08   0.33   2.87   4.63   3.99   4.10   4.29   0.98   0.05  

Operational  Vitality  

3.86   0.39   2.56   4.53   3.69   3.90   4.14   0.98   0.06  

Total  Score   4.08   0.30   3.12   4.56   3.98   4.09   4.27   0.99   0.03  

The  composite  scale  and  subscale  scores  exhibited  excellent  reliability,  with  all  coefficients  estimated  >  .98.  This  is  largely  due  to  the  Catholic  School  Program  Effectiveness  Survey  of  Adults  having  a  large  number  of  highly  correlated  items,  each  being  rated  by  several  respondents  per  school.  Based  on  an  estimate  of  the  standard  deviation  for  the  composite  (scale)  and  subscale  scores  and  their  respective  reliability  coefficients,  

the  standard  errors  of  measurement  can  be  calculated  as .  The  standard  error  of  

measurement  can  be  used  to  form  confidence  bands  around  scores  for  specific  schools.  It  should  be  noted  that  while  the  exceptionally  high  reliability  for  this  scale  is  a  positive  attribute,  achievement  of  the  high  reliability,  in  large  part,  comes  at  the  cost  of  a  relatively  lengthy  instrument.  In  the  psychometric  literature,  reliabilities  of  .90  are  often  considered  sufficient  for  most  or  all  practical  uses  of  an  instrument.  With  reliabilities  generally  approaching  or  exceeding  .99,  one  or  more  items  could  be  trimmed  from  most  subscales  while  retaining  sufficient  reliability  and  breadth  of  coverage.  This  might  facilitate  more  efficient  survey  administration  and  less  response  burden.    

 Observed  subscale  correlations  are  presented  in  Table  5.  Correlation  Matrix  for  Subscale  Scores  (School-­‐level)  Derived  from  the  Catholic  School  Program  Effectiveness  Survey  of  Adults.  These  correlations  are  very  high,  with  an  average  correlation  of  .85,  and  indicate  substantial  overlap  in  the  information  conveyed  by  

Page 16: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  16      

the  subscale  scores  and  support  a  one-­‐factor  conceptualization  and  use  of  the  Catholic  School  Program  Effectiveness  Survey  of  Adults.    

Table  5.  Correlation  Matrix  for  Subscale  Scores  (School-­level)  Derived  from  the  Catholic  School  Program  Effectiveness  Survey  of  Adults  

 Mission  and  

Catholic  Identity  

Governance  and  

Leadership  

Academic  Excellence  

Operational  Vitality  

Mission  and  Catholic  Identity  

-­‐-­‐        

Governance  and  Leadership  

0.90   -­‐-­‐      

Academic  Excellence  

0.86   0.89   -­‐-­‐    

Operational  Vitality  

0.77   0.85   0.81   -­‐-­‐  

Note:  All  correlations  are  statistically  significant  at  p  <  .001.  

 

Summary  and  Recommendations  

Based  on  the  results  of  these  analyses,  the  Catholic  School  Program  Effectiveness  Survey  of  Adults  appears  to  measure  a  single  construct  or  factor  reflecting  the  general  shared  perceptions  of  schools’  adherence  to  the  Standards.  All  items  were  supported  as  valid  and  reliable  indicators  of  a  general  Standards  construct,  and  reliability  estimates  of  a  composite  score  computed  by  averaging  over  respondents  and  items  within  the  same  school  indicated  exceptionally  high  reliability  (i.e.,  very  minimal  measurement  error).  In  fact,  the  attained  degree  of  reliability  affords  some  opportunity  to  shorten  the  scale  by  trimming  items  that  exhibit  high  rates  of  missing  data  (skipped  questions  and  Don’t  Know  responses)  to  ease  respondent  burden  while  maintaining  excellent  reliability.      Perhaps  contrary  to  initial  expectation,  there  was  no  evidence  that  this  survey  instrument  measures  four  distinct  factors  relating  to  the  four  domains  articulated  in  the  Standards  (viz.,  Mission  and  Catholic  Identity,  Governance  and  Leadership,  Academic  Excellence,  and  Operational  Vitality).  This  finding  has  one  or  more  explanations,  each  with  potentially  important  implications  for  the  interpretation  and  use  of  the  survey.  These  four  theoretical  constructs/domains  may  be  ecologically  valid  and  distinct,  but  could  be  highly  correlated  in  that  schools  which  tend  do  well  in  one  domain  will  also  tend  do  well  in  the  other  domains.  This  may  be  particularly  true  for  the  sample  of  schools  that  self-­‐selected  to  participate  in  this  psychometric  study.  Although  psychometric  concepts  such  as  validity  and  reliability  are  often  ascribed  to  instruments,  they  are  more  accurately  considered  properties  of  the  intended  inferences  or  interpretations  made  from  measurement  scores,  which  includes  not  only  the  instrument  as  stimulus  but  also  the  characteristics  of  the  respondent  population,  conditions  during  measurement,  and  inferences  made  on  the  basis  of  the  scores.  It  may  be  that  regardless  of  item  construction,  the  respondents  in  this  sample  may  not  cognitively  distinguish  amongst  the  domains.  That  is,  although  the  domains  were  developed  based  on  a  considerable  body  of  theoretical  and  empirical  literature,  this  theoretical  and  empirical  knowledge  may  not  be  sufficiently  developed  within  the  majority  of  respondents  for  them  to  differentiate  amongst  the  domains  in  their  individual  or  collective  responses  to  the  survey.  Without  this  knowledge,  discernment  amongst  the  domains  by  the  respondents  may  be  unrealistic,  regardless  of  the  survey  items.  Even  in  the  absence  of  such  

Page 17: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  17      

exposure,  all  of  this  is  not  to  suggest  that  schools  should  not  consider  computing  and  interpreting  subscale  scores.  Examination  of  subscale  scores  might  be  useful,  for  instance,  if  a  school  is  evaluating  a  concerted  effort  to  improve  its  efforts  with  respect  to  a  particular  standard/domain.  It  might  that  such  a  concerted  effort  targeting  improvements  pertaining  to  a  single  domain  might  generate  movement  unique  to  the  subscale  measuring  the  targeted  domain.  It  is  also  possible  that  schools  that  may  not  score  uniformly  across  the  subscales  may  be  underrepresented  in  this  present  sample.  If  such  is  the  case,  the  subscales  might  be  of  more  individual  importance  for  these  schools.  However,  in  general,  the  subscale  scores  corresponding  to  each  of  the  domains  would  be  expected  to  be  highly  correlated  with  one  another  and  thus  convey  little  unique  information.  In  such  cases,  focusing  on  the  total  composite  score  would  provide  a  more  reliable,  albeit  general,  measure  of  adherence  to  the  Standards.    Certain  items,  particularly  those  from  the  Operational  Vitality  or  Academic  Excellence  subscales,  exhibited  high  rates  of  either  being  skipped  by  the  respondent  or  receiving  a  Don’t  Know  response.  Based  on  the  content  of  the  items  and  pattern  of  Don’t  Know  responses,  it  seems  likely  that  the  respondents,  particularly  non-­‐school-­‐based  adults,  to  the  survey  may  not  have  adequate  knowledge  of  the  content  being  tapped  by  these  items.  To  the  extent  that  these  items  require  knowledge  of  their  school  that  the  respondents  may  not  have,  some  of  the  respondents  who  did  respond  to  the  item  may  have  done  so  with  little  actual  knowledge  and  either  made  a  “guess”  or  answered  in  a  way  to  be  consistent  with  their  responses  on  other  items  or  overall  impression  of  their  school.  These  items  should  be  reviewed  and  considered  for  deletion  from  the  survey.13  It  may  be  possible  to  re-­‐write  some  items  so  as  to  be  clearer  to  respondents.  It  would  be  desirable  to  have  missing  data  rates  less  than  10%.  The  worst  offending  items  with  the  rates  of  missing  data  greater  than  30%  are  listed  below:  

Our  school’s  financial  plan  is  the  result  of  a  collaborative  process  including  expert  advisors.   Our  school  has  an  institutional  advancement  plan,  based  on  our  mission,  which  uses  current  and  

effective  strategies  for  communications,  marketing,  enrollment  management,  and  development.   Our  school  leaders  take  responsibility  for  ensuring  that  the  financial  plans  and  budgets  are  

implemented  using  best  practices.   Our  school  maintains  and  shares  a  technology  management  plan.   Our  school  treats  all  personnel  with  consistency,  fairness,  and  justice.   Our  school  maintains  and  shares  plans  for  managing  facilities  and  equipment.   At  our  school,  all  administrators,  faculty  and  staff  engage  in  ongoing  professional  development.   Teachers  vary  the  types  of  assessments  they  use  to  monitor  individual  and  class-­‐wide  student  

learning.  

The  above  items  were  associated  with  the  highest  rates  of  missing  data.  But  other  items  had  higher  than  desired  rates  of  missing  data  as  well.  These  items  tended  to  be  towards  the  latter  half  of  the  surveys  and  may  suggest  respondent  fatigue  or  disinterest  in  the  survey.  Trimming  the  scale  of  items  that  exhibited  the  highest  non-­‐response  rates  may  ease  the  response  burden  and  address  this  issue.  It  might  also  be  prudent  not  to  group  subscale  items  together.  Currently,  the  Academic  Excellence  and  Operational  Vitality  subscales,  which  are  in  the  latter  half  of  the  survey,  experience  a  disproportionate  share  of  missing  data.  It  

                                                                                                               

13  Note  that  deleting  reliable  items  can  reduce  subscale  reliability.  This  is  not  as  much  of  a  concern  for  scales  with  many  items,  but  may  be  for  scales  with  fewer  than  five  to  seven  items.  If  reliability  becomes  a  concern  after  deleting  items,  one  can  attempt  to  address  this  issue  by  generating  replacement  items  that  measure  the  construct  of  interest  for  those  that  are  deleted.  

Page 18: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  18      

may  be  that  these  items/subscales  are  less  familiar  to  respondents  or  it  may  be  they  are  fatigued  or  bored  by  the  time  they  reach  these  items.  Any  missingness  due  to  fatigue/boredom  may  be  more  evenly  distributed  across  subscales  if  a  random  mixing  of  items  was  employed.    A  5-­‐point  Likert  scale  was  employed  with  anchors  provided  for  only  the  extreme  points  (Strongly  Disagree,  Strongly  Agree).  Adding  labels  for  the  intermediate  points  may  (though  not  necessarily)  increase  the  range  of  the  scale  being  used  by  the  respondents,  which  may  translate  into  more  variability  in  ratings  and  subscale  scores  across  schools.  To  the  extent  that  the  increased  variability  is  reliable  variation,  adding  labels  for  all  points  would  tend  to  increase  reliability.  It  might  also  help  standardize  the  cognitive  mapping  used  by  respondents  to  link  scale  points  to  internal  sense  of  agreement  with  the  items.  This  may  increase  reliability  as  random  individual  differences  in  this  mapping  with  the  current  labeling  scheme  may  manifest  as  measurement  error.  Importantly,  using  standard  labels  (e.g.,  Strongly  Disagree,  Disagree,  Neither  Agree/Disagree,  Agree,  Strongly  Agree)  may  help  achieve  item  and  scale  scores  that  better  approximate  interval-­‐level  properties  and  do  so  more  consistently  so  across  respondents.  Ordinal  scales  only  convey  information  about  the  rank,  whereas  interval-­‐level  measures  also  convey  the  amount  of  difference.    Finally,  a  few  cautionary  statements  pertaining  to  these  analyses  are  necessary.  These  analyses  were  conducted  with  less  than  optimal  number  of  schools  for  assessing  the  school-­‐level  psychometric  properties  of  the  survey.  Although  the  large  number  of  respondents  within  each  participating  school  and  the  generally  high  reliability  (and  factor  loadings)  of  most  items  facilitated  the  analysis,  it  would  have  been  desirable  to  have  at  least  150-­‐200  schools.  Related  and  also  impinging  on  the  results  is  that  the  high  non-­‐response  rate  of  schools  introduces  the  possibility  of  self-­‐selection  effects.  The  implication  of  this  is  that  the  obtained  sample  may  not  be  representative  of  the  broader  population  of  schools.  Therefore,  the  results  may  not  fully  generalize  to  that  broader  population.                  

Page 19: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  19      

 Student  Survey  (Grades  5-­8)  

Respondent  Characteristics  

A  total  of  4,343  respondents  from  46  schools  (average  of  94  respondents  per  school)  completed  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­‐  8th  grades)  for  this  psychometric  study.  Participating  students  were  nearly  evenly  distributed  across  grade  levels  (21.2%  to  28.5%  in  the  four  grades).  Just  over  half  of  respondents  were  female  (55%),  and  the  racial  composition  of  the  sample  was:  White  (81.5%),  African  American  or  Black  (5.7%),  multi-­‐racial  (7.1%),  Asian  (4.4%),  American  Indian  or  Alaska  Native  (.9%),  or  Native  Hawaiian  or  Pacific  Islander  (.4%).  The  majority  of  respondents  identified  as  non-­‐Hispanic  (92.4%).    Most  respondents  expressed  their  religious  affiliation  as  Catholic  (87.1%).  Most  respondents  also  reported  being  affiliated  with  the  school  for  more  than  4  years  (71.1%).    

Response  Pattern  Summaries  

Response  frequencies  and  intraclass  correlation  coefficients  (ICCs)  for  each  of  the  27  items  are  reported  in  Table  A.4.  Item  Response  Frequencies  and  Intraclass  Correlation  Coefficients  (ICC)  for  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5-­‐8  Grades)  in  Appendix  AAppendix  .  On  9  of  the  items,  at  least  75%  of  middle  school  students  endorsed  the  Strongly  Agree  or  adjacent  response  options,  and  all  items  were  endorsed  (Strongly  Agree  or  adjacent  response)  by  at  least  50%  of  respondents.  Though  response  frequencies  are  clustered  towards  endorsement  of  options  indicating  greater  agreement,  there  was  some  degree  of  variability  of  frequency  patterns  across  items.  Item  percentages  for  skipped  or  Don't  Know  responses  ranged  between  4.7%  and  21.9%  (Mean  =  9.5%).  One  item  was  skipped  or  had  a  Don't  Know  response  from  20%  or  more  of  respondents  (viz.,  At  our  school,  teachers  use  student  work  and  student  test  results  to  improve  how  they  teach.).  Of  note  is  that  the  rate  of  using  the  Don’t  Know  option  was  much  higher  for  this  item  than  all  other  items,  which  may  indicate  that  respondents  did  not  feel  they  had  sufficient  knowledge  to  provide  their  rating  for  this  item.  Three  other  items  had  a  Don’t  Know  response  rate  greater  than  10%.  However  explained,  these  items  should  be  considered  for  rewording,  replacement  or  deletion  from  the  scale,  along  with  other  items  exhibiting  the  highest  skip  or  Don't  Know  rates,  given  the  problems  missing  data  presents  for  estimation  of  subscale  and  scale  scores.14    ICCs  quantify  the  proportion  of  variability  in  responses  that  is  attributable  to  variability  between  scores.  In  other  terms,  an  ICC  reflects  the  degree  of  non-­‐independence  amongst  responses  from  students  affiliated  with  the  same  school,  with  0  =  independence  (i.e.,  no  systematic  variation  across  schools)  and  1  =  complete  dependence  (i.e.,  all  variation  in  responses  is  due  to  differences  across  schools).  As  this  survey  is  intended  to  measure  school-­‐level  adherence  to  Standards,  ICCs  greater  than  zero  are  to  be  expected  and  desired.  Additionally,  ICCs  greater  than  .01  support  the  need  for  statistical  methods  that  account  for  the  observed  non-­‐independence.  The  ICCs  for  most  items  were  moderate  in  magnitude,  reflecting  similarity  in  responses  

                                                                                                               

14  Although  most  analyses  described  in  the  Factorial  Validity  and  Reliability  use  statistical  procedures  that  can  incorporate  respondents  that  are  missing  data  on  one  or  more  items,  respondents  missing  data  on  all  items  are  necessarily  omitted  from  these  analyses.  Depending  on  the  particular  analysis,  between  138  and  211  respondents  were  omitted  from  subsequent  analysis  due  to  missing  data  on  all  items.  

Page 20: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  20      

from  students  affiliated  the  same  school  and  justifying  the  need  for  statistical  methods  that  can  account  for  non-­‐independence  amongst  the  observations.  ICCs  across  all  items  ranged  from  .02  to  .21  (Mean  =  .07),  indicating  that  between  2%  and  21%  of  variation  in  item  responses  was  attributable  to  respondents  being  affiliated  with  different  schools.  

The  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­‐  8th  grades)  is  intended  to  provide  data  on  Catholic  schools’  adherence  to  the  Standards.  Given  this  intent,  the  focus  of  these  analyses  is  on  school-­‐level  aggregations  of  the  responses.  Table  A.5.  School-­‐level  Statistics  for  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5-­‐8  Grades)  in  Appendix  A:  Tables  from  the  Item-­‐level  Analyses  of  the    provides  statistics  to  describe  the  distributions  of  these  school-­‐level  aggregates,  where  student  responses  for  each  item  are  averaged  with  other  students  from  their  school.  The  theoretical  range  for  these  aggregated  item  averages  is  1  (Strongly  Disagree)  -­‐  5  (Strongly  Agree),  with  higher  scores  indicating  higher  within-­‐school  average  agreement  for  the  particular  item.  On  average,  schools  were  rated  highly  on  all  of  the  items  (mean  average  rating  across  items  =  4.01).  Along  with  average,  minimum,  and  maximum  ratings,  standard  deviations  and  quartiles  are  also  provided  for  each  item.  The  median  (also  known  as  the  50th  percentile  or  second  quartile)  reflects  the  item  score  at  which  50%  of  schools  fall  at  or  below.  For  all  items,  50%  or  more  of  the  schools  scored  at  least  a  3.53.  Similarly  to  the  respondent-­‐level  item  statistics,  scores  were  clustered  toward  the  upper  end  of  the  distribution.  

Factorial  Validity  and  Reliability  

Multilevel,  ordinal  confirmatory  factor  analyses  (CFA)  began  with  the  separate  examination  of  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­‐  8th  grades)  subscales  corresponding  to  each  of  the  four  Standards  for  Catholic  School  Program  Effectiveness.  These  CFA  models,  consisting  of  one-­‐factor  with  three  to  twelve  indicators  each,  were  fit  to  the  data.  Then,  these  models  were  combined  into  a  single,  four-­‐factor  CFA  model  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­‐  8th  grades).  CFA  concluded  with  estimation  of  a  model  where  all  items  were  specified  as  measures  of  a  single  factor  (one-­‐factor  CFA  model).15  The  model  fit  tests  and  indices  for  these  analyses  are  reported  in  Table  6.  Model  Fit  Tests  and  Indices  for  Full-­‐Scale  and  Subscale  CFA  Models  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­‐  8th  grades)Multilevel,  ordinal  confirmatory  factor  analyses  (CFA)  began  with  the  separate  examination  of  Catholic  School  Program  Effectiveness  Survey  of  Adults  subscales  corresponding  to  each  of  the  four  Standards  for  Catholic  School  Program  Effectiveness.  These  CFA  models,  consisting  of  one-­‐factor  with  seven  to  fourteen  indicators  each,  were  fit  to  the  data.  Then,  these  models  were  combined  into  a  single,  four-­‐factor  CFA  model  of  the  Catholic  School  Program  Effectiveness  Survey  of  Adults.  CFA  concluded  with  estimation  of  a  model  where  all  items  were  specified  as  measures  of  a  single  factor  (one-­‐factor  CFA  model).  The  model  fit  tests  and  indices  for  these  analyses  are  reported  in  Table  1Error!  Not  a  valid  bookmark  self-­‐reference..  These  tests  evaluate  how  well  each  model  fits  the  data.  The  aim  is  to  identify  a  model  that  fits  the  data  well  (within  sampling  error)  or  approximately  well  and  is  parsimonious.  If  a  model  fits  the  data  poorly,  the  model  would  be  rejected  from  further  consideration.    

                                                                                                               

15  In  these  analyses,  model  parameters  are  being  simultaneously  estimated  at  two-­‐levels,  respondent-­‐level  and  the  school-­‐level.,  with  an  unstructured  model  specified  for  the  within-­‐level  (i.e.,  no  over-­‐identifying  constraints  included).  However,  the  focus  of  these  analyses  is  on  discovering  the  factors  that  pertain  to  schools.    Thus,  the  focus  of  this  report  is  exclusively  on  results  pertaining  to  the  school-­‐level  models.  

Page 21: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  21      

Table  .  These  tests  evaluate  how  well  each  model  fits  the  data.  The  aim  is  to  identify  a  model  that  fits  the  data  well  (within  sampling  error)  or  approximately  well  and  is  parsimonious.  If  a  model  fits  the  data  poorly,  the  model  would  be  rejected  from  further  consideration.                    

Table  6.  Model  Fit  Tests  and  Indices  for  Full-­Scale  and  Subscale  CFA  Models  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­  8th  grades)  

 Chi-­‐Square   df   p-­‐value   RMSEA   CFI   SRMR  

Mission  and  Catholic  Identity  

27.25   27   0.45   0.001   1.00   0.087  

Governance  and  Leadershipa   -­‐-­‐   -­‐-­‐   -­‐-­‐   -­‐-­‐   -­‐-­‐   -­‐-­‐  

Academic  Excellence   49.82   54   0.64   <0.001   1.00   0.101  

Operational  Vitalitya   -­‐-­‐   -­‐-­‐   -­‐-­‐   -­‐-­‐   -­‐-­‐   -­‐-­‐  

1-­‐Factor  Model   333.83   324   0.34   0.003   1.00   0.126  

4-­‐Factor  Model   324.23   318   0.39   0.002   1.00   0.122  

Note:  df  =  degrees  of  freedom;  RMSEA  =  Root  Mean  Square  Error  of  Approximation;  CFI  =  Comparative  Fit  Index;  SRMR  =  Standardized  Root  Mean  Residual.  aWith  only  3  items  for  each  of  these  subscales,  these  models  are  completely  saturated  (df=0)  and  thus  no  meaningful  model  fit  tests  or  indices  are  available.    

According  to  the  conventional  guidelines  described  earlier  in  the  methodology  section,  all  subscale  models  met  exact  fit  standards.  The  four-­‐factor  CFA  model,  which  combined  the  subscale  models  into  a  single  analysis,  similar  exhibited  good  model  fit.  However,  factor  inter-­‐correlations  were  notably  very  high  (.80  through  1.093),16  indicating  substantial  overlap  and  little  discriminant  validity  amongst  the  factors.  Therefore,  a  one-­‐factor  CFA  model  was  examined.  As  reflected  by  the  model  fit  indices,  the  one-­‐factor  model  fit  the  data  according  to  exact-­‐fit  standards.  Still,  it  is  essential  that  the  one-­‐factor  model  solution  be  considered  meaningful  and  interpretable  where  the  meaningfulness  or  interpretability  of  a  solution  is  determined  by  considering  the  strength  and  pattern  of  relationships  between  the  items  and  underlying  (latent)  factor.  The  relationships  between  factors  and  indicators  are  typically  depicted  in  a  factor  loading  matrix.  Standardized  factor  loadings  and  associated  standard  errors  for  the  one-­‐factor  CFA  model  and  for  

                                                                                                               

16  Correlations  are  bounded  by  the  interval  [-­‐1,1].  The  model-­‐estimated  correlation  between  the  Vitality  and  Academic  Excellence  factors  exceeded  1  and  is  a  non-­‐admissible  value.  This  likely  stems  from  sampling  and  minor  specification  error  in  combination  with  a  high  population  correlation  between  the  factors.  

Page 22: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  22      

each  one-­‐factor  subscale  model  are  provided  in  Table  in  Appendix  A:  Tables  from  the  Item-­‐level  Analyses  of  the  .  The  standardized  factor  loadings  were  mostly  high,  with  the  vast  majority  greater  than  0.7  and  all  but  two  loadings  above  0.48,  indicating  that  these  items  are  salient  and  highly  reliable  measures  of  the  factor.  Two  items,  however,  had  marginal  factor  loadings:  (1)  Our  school  offers  guidance  and  resources  (such  as  counselors,  tutors  and  special  teachers)  to  help  students  be  successful  and  (2)  Our  school  offers  programs  and  activities  (such  as  sports,  drama,  clubs,  or  band)  for  students  to  develop  their  gifts  and  talents.  Only  approximately  10%  of  the  variation  in  the  items  is  explained  by  variability  in  the  underlying  factor,  and  the  items  were  noticeably  poorer  measures  of  the  latent  factor  compared  to  other  items.    

Though  the  CFA  analyses  and  higher  inter-­‐subscale  correlations  suggest  a  one-­‐factor  model,  it  is  possible  that  another  k-­‐factor  model  not  examined  generated  the  data.  Therefore,  an  exploratory  factor  analysis  (EFA)  was  performed  prior  to  settling  on  the  one-­‐factor  model  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­‐  8th  grades).  An  important  and  initial  task  in  conducting  exploratory  factor  analysis  is  to  determine  the  number  of  factors  or  dimensions  that  are  being  measured  by  the  survey  instrument.  This  determination  is  guided  by  statistical  tests  and  indices,  evaluation  of  eigenvalues,  and  meaningfulness  and  interpretability  of  the  solution.  As  with  the  confirmatory  factor  analyses,  there  are  two-­‐levels  under  consideration,  respondent-­‐level  and  the  school-­‐level,  and  the  number  of  factors  would  typically  be  determined  at  each  level.  However,  the  focus  of  this  report  is  on  the  school  as  the  unit  of  analysis.  

Eigenvalues  quantify  the  variance  in  the  item  responses  that  is  explained  by  the  factors.  Factors  that  account  for  more  variation  are  considered  potentially  more  important  or  meaningful  than  factors  that  account  for  less  variation.  Eigenvalues  from  the  school-­‐level  factor  analysis  of  the  survey  data  are  reported  in  Table  7.  Eigenvalues  for  the  Exploratory  Factor  Analysis  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­‐  8th  grades).  An  often  cited  rule  of  thumb  is  the  Kaiser  rule,  which  states  that  factors  with  eigenvalues  greater  than  1.0  should  be  extracted.  According  to  the  Kaiser  rule,  a  model  with  five  factors  should  be  examined.  This  rule,  however,  has  been  criticized  as  leading  to  extraction  of  too  many  factors  and  thus  typically  should  be  considered  an  upper  bound  estimate  of  the  number  of  factors.      

Table  7.  Eigenvalues  for  the  Exploratory  Factor  Analysis  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­  8th  grades)  

    Factor   1   2   3   4   5   6   7   8   9   10   ...  

Eigenvalues       16.26   2.51   2.19   1.55   1.31   0.91   0.64   0.54   0.47   0.46    

 Another  factor  enumeration  approach  used  is  examination  of  the  Scree  plot,  depicted  in  Figure  2.  Scree  Plot  of  Eigenvalues  (School-­‐Level)  for  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­‐  8th  grades).  With  this  approach,  one  typically  seeks  the  point  where  there  is  a  pronounced  bend  (elbow)  in  the  curve.  Factors  before  the  bend  are  given  further  consideration.  According  to  the  scree  plot,  a  one-­‐factor  or  perhaps  a  two-­‐factor  model  should  be  considered  for  extraction.  Although  more  sophisticated  methods  for  factor  enumeration  exists  (e.g.,  parallel  analysis),  these  were  computationally  infeasible  or  inaccessible  for  the  present  analysis,  which  involve  a  multilevel  structure  and  ordinal  item  responses  distributions  that  make  it  difficult  to  employ  more  advanced  methods.  Thus,  the  one-­‐factor  and  two-­‐factor  models  were  given  closer  consideration.  

Page 23: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  23      

Figure  2.  Scree  Plot  of  Eigenvalues  (School-­Level)  for  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­  8th  grades)  

   In  conjunction  with  examination  of  the  eigenvalues  and  scree  plot,  statistical  tests  and  indices  of  model  fit  are  often  consulted.  All  models  fit  the  data  well  (see  Table  8.  Model  Fit  Tests  and  Indices  for  EFA  Models  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­‐  8th  grades)).  With  most  indices  being  within  desired  ranges  for  all  models  under  consideration,  the  philosophical  principle  known  as  Occam's  razor  dictates  that  one  would  choose  the  most  parsimonious  model,  which  often  is  the  model  with  the  fewest  factors.  Consistent  with  the  CFA,  model  fit  indices,  in  conjunction  with  the  eigenvalues,  scree  plot  and  factor  pattern  matrix,  suggested  a  model  that  posits  a  single  factor  for  explaining  and  summarizing  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­‐  8th  grades)  item  responses  when  aggregated  to  the  school-­‐level.  This  conclusion  was  corroborated  by  the  less  interpretable  solutions  for  the  EFA  models  with  two  or  more  factors  (e.g.,  factor  loadings  >  1,  most  items  load  on  a  general  factor,  with  some  Academic  Excellence  and  other  items  cross-­‐loading  onto  a  second  factor).  

Table  8.  Model  Fit  Tests  and  Indices  for  EFA  Models  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­  8th  grades)  

 1-­‐Factor  Model  

2-­‐Factor  Model  

3-­‐Factor  Model  

4-­‐Factor  Model  

5-­‐Factor  Model  

Chi-­‐Square   333.83   291.00   248.85   215.15   185.94  

df   324   298   273   249   226  

p-­‐value   0.34   0.60   0.85   0.94   0.98  

RMSEA   0.003   <0.001   <0.001   <0.001   <0.001  

CFI   1.00   1.00   1.00   1.00   1.00  

SRMR   0.126   0.102   0.080   0.064   0.049  

0  

2  

4  

6  

8  

10  

12  

14  

16  

18  

1   2   3   4   5   6   7   8   9   10  

Eigenvalue

 

Factor  

Page 24: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  24      

 Summary  statistics  for  the  distribution  of  composite  scores  formed  by  taking  the  average  of  school-­‐level  item  averages  for  the  entire  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­‐  8th  grades)  and  each  subscale  can  be  found  in  Table  9.  Descriptive  Statistics  for  a  Total  Scale  Composite  Score  and  Subscale  Scores  (School-­‐level)  Derived  from  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­‐  8th  grades).  The  theoretical  range  for  the  composite  and  subscale  scores  is  1  -­‐  5,  with  5  reflecting  greater  adherence  to  the  Standards  as  reported  by  students.  On  average,  schools  tended  to  score  highly  on  the  scale  and  each  subscales,  with  75%  or  more  schools  scoring  greater  than  3.67  on  all  scales.    

Table  9.  Descriptive  Statistics  for  a  Total  Scale  Composite  Score  and  Subscale  Scores  (School-­level)  Derived  from  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­  8th  grades)  

 

Mean Standard Deviation Minimum Maximum 25th

Percentile Median 75th Percentile Reliability SEM

Mission  and  Catholic  Identity  

4.06   0.22   3.52   4.49   3.92   4.08   4.23   0.93   0.06  

Governance  and  Leadership  

4.12   0.45   2.71   4.85   3.92   4.22   4.44   0.99   0.05  

Academic  Excellence  

3.97   0.28   3.19   4.56   3.87   4.00   4.19   0.88   0.10  

Operational  Vitality  

3.90   0.29   3.20   4.44   3.67   3.92   4.11   0.78   0.14  

Total  Score   4.01   0.26   3.36   4.51   3.92   4.01   4.23   0.96   0.05  

The  composite  scale  and  subscale  scores  exhibited  excellent  reliability,  except  for  the  Operational  Vitality  scale.  The  reliability  estimate  for  the  Operational  Vitality  subscale  was  marginally  acceptable  (.78).  Based  on  an  estimate  of  the  standard  deviation  for  the  composite  (scale)  and  subscale  scores  and  their  respective  

reliability  coefficients,  the  standard  errors  of  measurement  can  be  calculated  as .  The  

standard  error  of  measurement  can  be  used  to  form  confidence  bands  around  scores  for  specific  schools.  It  should  be  noted  that  while  the  high  reliability  for  this  scale  is  a  positive  attribute,  achievement  of  the  high  reliability,  in  large  part,  comes  at  the  cost  of  a  relatively  lengthy  instrument.  In  the  psychometric  literature,  reliabilities  of  .90  are  often  considered  sufficient  for  most  or  all  practical  uses  of  an  instrument.  With  reliabilities  generally  exceeding  .93  on  three  subscales,  one  or  more  items  could  be  trimmed  from  these  subscales  while  retaining  sufficient  reliability  and  breadth  of  coverage.17  This  might  facilitate  more  efficient  survey  administration  and  less  response  burden.    

 Observed  subscale  correlations  are  presented  in  Table  10.  Correlation  Matrix  for  Subscale  Scores  (School-­‐level)  Derived  from  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­‐  8th  grades).  The  

                                                                                                               

17  Due  to  the  Governance  and  Leadership  subscale  being  comprised  of  just  three  items,  it  is  not  recommended  that  items  be  trimmed  from  this  subscale.  

Page 25: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  25      

correlations  are  smaller  than  the  factor  inter-­‐correlations  reported  earlier  as  these  may  be  attenuated  by  measurement  error  and  are  obtained  using  a  different  estimator.  However,  these  correlations  were  also  very  high,  with  an  average  correlation  of  .76.  These  statistics  indicate  substantial  overlap  in  the  information  conveyed  by  the  subscale  scores  and  support  a  one-­‐factor  conceptualization  and  use  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­‐  8th  grades).    

Table  10.  Correlation  Matrix  for  Subscale  Scores  (School-­level)  Derived  from  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­  8th  grades)  

 Mission  and  

Catholic  Identity  

Governance  and  

Leadership  

Academic  Excellence  

Operational  Vitality  

Mission  and  Catholic  Identity  

-­‐-­‐        

Governance  and  Leadership  

0.73   -­‐-­‐      

Academic  Excellence  

0.83   0.80   -­‐-­‐    

Operational  Vitality  

0.69   0.64   0.84   -­‐-­‐  

Note:  All  correlations  are  statistically  significant  at  p  <  .001.  

 

Summary  and  Recommendations  

Based  on  the  results  of  these  analyses,  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­‐  8th  grades)  appears  to  measure  a  single  construct  or  factor  reflecting  the  general  shared  perceptions  of  schools’  adherence  to  the  Standards.  Most  items  were  supported  as  valid  and  reliable  indicators  of  a  general  Standards  construct,  and  reliability  estimates  of  a  composite  score  computed  by  averaging  over  respondents  and  items  within  the  same  school  indicated  exceptionally  high  reliability  (i.e.,  very  minimal  measurement  error)  for  all  constructs  except  Operational  Vitality.  The  reliability  estimate  for  Operational  Vitality  was  marginal,  which  is  partially  attributable  to  this  scale  having  few  (3)  items.  Adding  additional,  reliable  and  valid  items  to  measure  Operational  Vitality  would  likely  increase  this  subscale’s  reliability.  For  the  other  subscale  scores  and  the  total  score,  the  attained  degree  of  reliability  affords  some  opportunity  to  shorten  the  scale  by  trimming  two  items  that  exhibited  relatively  lower  factor  loadings  and  other  items  that  exhibit  high  rates  of  missing  data  (skipped  questions  and  Don’t  Know  responses).  Trimming  select  items  could  ease  respondent  burden  while  maintaining  or,  in  the  case  of  the  two  items  with  low  factor  loadings,  increasing  reliability.      Perhaps  contrary  to  initial  expectation,  there  was  no  evidence  that  this  survey  instrument  measures  four  distinct  factors  relating  to  the  four  domains  articulated  in  the  Standards  (viz.,  Mission  and  Catholic  Identity,  Governance  and  Leadership,  Academic  Excellence,  and  Operational  Vitality).  This  finding  has  one  or  more  explanations,  each  with  potentially  important  implications  for  the  interpretation  and  use  of  the  survey.  These  four  theoretical  constructs/domains  may  be  ecologically  valid  and  distinct,  but  could  be  highly  correlated  in  that  schools  which  tend  do  well  in  one  domain  will  also  tend  do  well  in  the  other  domains.  This  may  be  particularly  true  for  the  sample  of  schools  that  self-­‐selected  to  participate  in  this  psychometric  study.  Although  psychometric  concepts  such  as  validity  and  reliability  are  often  ascribed  to  instruments,  

Page 26: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  26      

they  are  more  accurately  considered  properties  of  the  intended  inferences  or  interpretations  made  from  measurement  scores,  which  includes  not  only  the  instrument  as  stimulus  but  also  the  characteristics  of  the  respondent  population,  conditions  during  measurement,  and  inferences  made  on  the  basis  of  the  scores.  It  may  be  that  regardless  of  item  construction,  the  respondents  in  this  sample  may  not  cognitively  distinguish  amongst  the  domains.  That  is,  although  the  domains  were  developed  based  on  a  considerable  body  of  theoretical  and  empirical  literature,  this  theoretical  and  empirical  knowledge  may  not  be  sufficiently  developed  within  the  majority  of  respondents  for  them  to  differentiate  amongst  the  domains  in  their  individual  or  collective  responses  to  the  survey.  Without  this  knowledge,  discernment  amongst  the  domains  by  the  respondents  may  be  unrealistic,  regardless  of  the  survey  items.  Even  in  the  absence  of  such  exposure,  all  of  this  is  not  to  suggest  that  schools  should  not  consider  computing  and  interpreting  subscale  scores.  Examination  of  subscale  scores  might  be  useful,  for  instance,  if  a  school  is  evaluating  a  concerted  effort  to  improve  its  efforts  with  respect  to  a  particular  standard/domain.  It  might  that  such  a  concerted  effort  targeting  improvements  pertaining  to  a  single  domain  might  generate  movement  unique  to  the  subscale  measuring  the  targeted  domain.  It  is  also  possible  that  schools  that  may  not  score  uniformly  across  the  subscales  may  be  underrepresented  in  this  present  sample.  If  such  is  the  case,  the  subscales  might  be  of  more  individual  importance  for  these  schools.  However,  in  general,  the  subscale  scores  corresponding  to  each  of  the  domains  would  be  expected  to  be  highly  correlated  with  one  another  and  thus  convey  little  unique  information.  In  such  cases,  focusing  on  the  total  composite  score  would  provide  a  more  reliable,  albeit  general,  measure  of  adherence  to  the  Standards.    Two  items  exhibited  with  relatively  low  factor  loadings:  (1)  Our  school  offers  guidance  and  resources  (such  as  counselors,  tutors  and  special  teachers)  to  help  students  be  successful  and  (2)  Our  school  offers  programs  and  activities  (such  as  sports,  drama,  clubs,  or  band)  for  students  to  develop  their  gifts  and  talents.  These  items  thus  did  not  emerge  as  good  measures  of  the  Academic  Excellence  construct  or  the  overarching  latent  construct.  They  should  be  reviewed  and  considered  for  re-­‐wording  or  deletion  from  the  scale.  Alternatively,  if  they  are  considered  informative  as  stand-­‐alone  items,  they  may  be  retained,  but  should  not  be  averaged  or  summed  with  the  other  items  when  computing  subscale  or  total  scale  composite  scores.    Certain  items,  particularly  those  from  the  Academic  Excellence  subscale,  exhibited  high  rates  of  either  being  skipped  by  the  respondent  or  receiving  a  Don’t  Know  response.  Based  on  the  content  of  the  items  and  pattern  of  Don’t  Know  responses,  it  seems  likely  that  the  respondents  to  the  survey  may  not  have  adequate  knowledge  of  the  content  being  tapped  by  these  items.  To  the  extent  that  these  items  require  knowledge  of  their  school  that  the  respondents  may  not  have,  some  of  the  respondents  who  did  respond  to  the  item  may  have  done  so  with  little  actual  knowledge  and  either  made  a  “guess”  or  answered  in  a  way  to  be  consistent  with  their  responses  on  other  items  or  overall  impression  of  their  school.  These  items  should  be  reviewed  and  considered  for  deletion  from  the  survey.18  It  may  be  possible  to  re-­‐write  some  items  so  as  to  be  clearer  to  respondents.  It  would  be  desirable  to  have  missing  data  rates  less  than  10%.  The  worst  offending  items  with  the  rates  of  missing  data  greater  than  15%  are  listed  below:  

At  our  school,  teachers  use  student  work  and  student  test  results  to  improve  how  they  teach.   At  our  school,  teachers  work  together  to  help  each  other  become  better  teachers.  

                                                                                                               

18  Note  that  deleting  reliable  items  can  reduce  subscale  reliability.  This  is  not  as  much  of  a  concern  for  scales  with  many  items,  but  may  be  for  scales  with  fewer  than  five  to  seven  items.  If  reliability  becomes  a  concern  after  deleting  items,  one  can  attempt  to  address  this  issue  by  generating  replacement  items  that  measure  the  construct  of  interest  for  those  that  are  deleted.  

Page 27: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  27      

In  our  school,  adults  do  service  on  behalf  of  the  poor  and  those  in  need.   Our  school’s  mission  affects  everything  we  do.  

The  above  items  were  associated  with  the  highest  rates  of  missing  data.  But  other  items  had  higher  than  desired  rates  of  missing  data  as  well.      A  5-­‐point  Likert  scale  was  employed  with  anchors  provided  for  only  the  extreme  points  (Strongly  Disagree,  Strongly  Agree).  Adding  labels  for  the  intermediate  points  may  (though  not  necessarily)  increase  the  range  of  the  scale  being  used  by  the  respondents,  which  may  translate  into  more  variability  in  ratings  and  subscale  scores  across  schools.  To  the  extent  that  the  increased  variability  is  reliable  variation,  adding  labels  for  all  points  would  tend  to  increase  reliability.  It  might  also  help  standardize  the  cognitive  mapping  used  by  respondents  to  link  scale  points  to  internal  sense  of  agreement  with  the  items.  This  may  increase  reliability  as  random  individual  differences  in  this  mapping  with  the  current  labeling  scheme  may  manifest  as  measurement  error.  Importantly,  using  standard  labels  (e.g.,  Strongly  Disagree,  Disagree,  Neither  Agree/Disagree,  Agree,  Strongly  Agree)  may  help  achieve  item  and  scale  scores  that  better  approximate  interval-­‐level  properties  and  do  so  more  consistently  so  across  respondents.  Ordinal  scales  only  convey  information  about  the  rank,  whereas  interval-­‐level  measures  also  convey  the  amount  of  difference.    Finally,  a  few  cautionary  statements  pertaining  to  these  analyses  are  necessary.  These  analyses  were  conducted  with  less  than  optimal  number  of  schools  for  assessing  the  school-­‐level  psychometric  properties  of  the  surveys.  Although  the  large  number  of  respondents  within  each  participating  school  and  the  generally  high  reliability  (and  factor  loadings)  of  most  items  facilitated  the  analysis,  it  would  have  been  desirable  to  have  at  least  150-­‐200  schools.  Related  and  also  impinging  on  the  results  is  that  the  high  non-­‐response  rate  of  schools  introduces  the  possibility  of  self-­‐selection  effects.  The  implication  of  this  is  that  the  obtained  sample  may  not  be  representative  of  the  broader  population  of  schools.  Therefore,  the  results  may  not  fully  generalize  to  that  broader  population.      

 

Page 28: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  28      

 Student  Survey  (Grades  9-­12)  

Respondent  Characteristics  

A  total  of  114  respondents  from  4  schools  (average  of  29  respondents  per  school)  completed  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­‐  12th  grades)  for  this  psychometric  study.  Participating  students  were  mostly  in  the  9th  (43%)  or  11th  (30%)  grades.  Just  over  half  of  respondents  were  male  (54%),  and  the  racial  composition  of  the  sample  was:  White  (73.7%),  African  American  or  Black  (13.2%),  multi-­‐racial  (8.8%),  Asian  (1.8%),  American  Indian  or  Alaska  Native  (.9%),  or  Native  Hawaiian  or  Pacific  Islander  (1.8%).  The  majority  of  respondents  identified  as  non-­‐Hispanic  (95.6%).    Most  respondents  expressed  their  religious  affiliation  as  Catholic  (71.9%).  Respondents  were  fairly  uniformly  distributed  in  their  number  of  years  attending  the  school  (10.5%  -­‐  23.7%).    

Response  Pattern  Summaries  

Response  frequencies  for  each  of  the  35  items  are  reported  in  Table  A.7.  Item  Response  Frequencies  for  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9-­‐12  Grades)  in  Appendix  AAppendix  .  On  17  of  the  items,  at  least  75%  of  high  school  students  endorsed  the  Strongly  Agree  or  adjacent  response  options,  and  all  items  were  endorsed  (Strongly  Agree  or  adjacent  response)  by  at  least  62%  of  respondents.  Though  response  frequencies  are  clustered  towards  endorsement  of  options  indicating  greater  agreement,  there  was  some  degree  of  variability  of  frequency  patterns  across  items.  Item  percentages  for  skipped  or  Don't  Know  responses  ranged  between  0%  and  32.5%  (Mean  =  21.0%).  Eleven  items  were  skipped  or  had  a  Don't  Know  response  from  25%  or  more  of  respondents.  The  most  extreme  instances  were  items  23  (At  our  school,  teachers  work  together  to  help  each  other  become  better  teachers),  24  (At  our  school,  all  administrators,  faculty  and  staff  engage  in  ongoing  professional  development),  and  26  (At  our  school,  teachers  work  together  to  use  data  on  student  performance  to  review  and  improve  courses  and  instruction).  These  items  on  the  Operational  Vitality  subscale  and  other  items  in  this  survey  may  ask  respondents  about  aspects  which  they  have  little  or  no  knowledge.  These  items  tended  to  have  the  highest  rates  of  Don’t  Know  responses.  However  explained,  these  items  should  be  considered  for  rewording,  replacement  or  deletion  from  the  scale,  along  with  other  items  exhibiting  the  highest  skip  or  Don't  Know  rates,  given  the  problems  missing  data  presents  for  estimation  of  subscale  and  scale  scores.19  A  pattern  of  items  in  the  second-­‐half  of  the  survey  tending  to  exhibit  higher  skipped  or  Don’t  Know  response  rates  than  the  first-­‐half  suggests  that  respondents  may  have  become  have  felt  rushed  or  less  motivated  (e.g.,  bored,  fatigued)  as  they  worked  towards  completing  the  survey.    

The  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­‐  12th  grades)  is  intended  to  provide  data  on  Catholic  schools’  adherence  to  the  Standards.  Given  this  intent,  the  desired  focus  of  analysis  is  on  school-­‐level  aggregations  of  the  responses.  However,  too  few  high  schools  were  successfully  recruited  for  

                                                                                                               

19  Although  most  analyses  described  in  the  Factorial  Validity  and  Reliability  use  statistical  procedures  that  can  incorporate  respondents  that  are  missing  data  on  one  or  more  items,  respondents  missing  data  on  all  items  are  necessarily  omitted  from  these  analyses.  Depending  on  the  particular  analysis,  between  12  and  27  respondents  were  omitted  from  subsequent  analysis  due  to  missing  data  on  all  items.  

Page 29: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  29      

this  psychometric  study  to  permit  school-­‐level  analyses.  Therefore,  all  subsequent  analyses  report  on  analyses  at  the  respondent-­‐level.  In  interpreting  the  results  of  these  analyses,  understand  that  these  results  may  not  extend  to  the  school-­‐level  and  any  such  generalization  is  discouraged.20  

Factorial  Validity  and  Reliability  

Ordinal  confirmatory  factor  analyses  (CFA)  began  with  the  separate  examination  of  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­‐  12th  grades)  subscales  corresponding  to  each  of  the  four  Standards  for  Catholic  School  Program  Effectiveness.  These  CFA  models,  consisting  of  one-­‐factor  with  four  to  fourteen  indicators  each,  were  fit  to  the  data.  Then,  these  models  were  combined  into  a  single,  four-­‐factor  CFA  model  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­‐  12th  grades).  CFA  concluded  with  estimation  of  a  model  where  all  items  were  specified  as  measures  of  a  single  factor  (one-­‐factor  CFA  model).21  The  model  fit  tests  and  indices  for  these  analyses  are  reported  in  Table  11.  Model  Fit  Tests  and  Indices  for  Full-­‐Scale  and  Subscale  CFA  Models  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­‐  12th  grades)Multilevel,  ordinal  confirmatory  factor  analyses  (CFA)  began  with  the  separate  examination  of  Catholic  School  Program  Effectiveness  Survey  of  Adults  subscales  corresponding  to  each  of  the  four  Standards  for  Catholic  School  Program  Effectiveness.  These  CFA  models,  consisting  of  one-­‐factor  with  seven  to  fourteen  indicators  each,  were  fit  to  the  data.  Then,  these  models  were  combined  into  a  single,  four-­‐factor  CFA  model  of  the  Catholic  School  Program  Effectiveness  Survey  of  Adults.  CFA  concluded  with  estimation  of  a  model  where  all  items  were  specified  as  measures  of  a  single  factor  (one-­‐factor  CFA  model).  The  model  fit  tests  and  indices  for  these  analyses  are  reported  in  Table  1Error!  Not  a  valid  bookmark  self-­‐reference..  These  tests  evaluate  how  well  each  model  fits  the  data.  The  aim  is  to  identify  a  model  that  fits  the  data  well  (within  sampling  error)  or  approximately  well  and  is  parsimonious.  If  a  model  fits  the  data  poorly,  the  model  would  be  rejected  from  further  consideration.    Table  .  These  tests  evaluate  how  well  each  model  fits  the  data.  The  aim  is  to  identify  a  model  that  fits  the  data  well  (within  sampling  error)  or  approximately  well  and  is  parsimonious.  If  a  model  fits  the  data  poorly,  the  model  would  be  rejected  from  further  consideration.    

Table  11.  Model  Fit  Tests  and  Indices  for  Full-­Scale  and  Subscale  CFA  Models  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­  12th  grades)  

 Chi-­‐Square   df   p-­‐value   RMSEA   CFI  

Mission  and  Catholic  Identity  

163.16   65   <0.001   0.122   0.97  

Governance  and  Leadership  

0.25   2   0.88   <0.001   1.00  

Academic  Excellence   172.34   77   <0.001   0.118   0.97  

                                                                                                               

20  In  addition,  the  respondent-­‐level  sample  size  was  particularly  modest  for  these  analyses,  particularly  those  of  the  entire  scale.  Caution  is  warranted  when  interpreting  the  results  until  more  data  is  available.  21  In  these  analyses,  model  parameters  are  being  simultaneously  estimated  at  two-­‐levels,  respondent-­‐level  and  the  school-­‐level.,  with  an  unstructured  model  specified  for  the  within-­‐level  (i.e.,  no  over-­‐identifying  constraints  included).  However,  the  focus  of  these  analyses  is  on  discovering  the  factors  that  pertain  to  schools.    Thus,  the  focus  of  this  report  is  exclusively  on  results  pertaining  to  the  school-­‐level  models.  

Page 30: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  30      

Operational  Vitality   0.23   2   0.89   <0.001   1.00  

1-­‐Factor  Model   923.81   560   <0.001   0.080   0.96  

4-­‐Factor  Model   763.71   554   <0.001   0.061   0.97  

Note:  df  =  degrees  of  freedom;  RMSEA  =  Root  Mean  Square  Error  of  Approximation;  CFI  =  Comparative  Fit  Index.    

According  to  the  conventional  guidelines  described  earlier  in  the  methodology  section,22  all  subscale  models  except  for  the  Governance  and  Leadership  and  Operational  Vitality  models  did  not  meet  exact  fit  standards.  The  Mission  and  Catholic  Identity  and  Academic  Excellence  models  also  failed  to  meet  approximate  fit  standards  for  the  RMSEA,  though  they  did  meet  the  conventional  threshold  for  the  CFI.  This  discrepancy  between  the  RMSEA  and  CFI  is  due  to  differences  in  how  they  “define”  or  quantify  misfit  and  their  proclivity  to  positive  bias  in  small  sample  sizes.  The  CFI  essentially  compares  the  misfit  of  the  model  at  hand  with  the  misfit  of  a  null-­‐model  that  specifics  no  correlation  amongst  variables.  Because  the  items  are  highly  correlated,  the  models  examined,  which  all  allow  some  degree  of  correlation  amongst  the  items,  fits  much  better  than  the  null  model;  hence  the  high  CFI.  The  RMSEA  estimates  absolute  misfit  per  degree  of  freedom  in  the  model  and  is  positively  biased,  particularly  for  small  sample  sizes.  Taken  all  together,  the  validity  of  the  subscale  models  for  Mission  and  Catholic  Identity  and  for  Academic  Excellence  is  suspect.  In  contrast,  the  subscale  models  for  Operational  Vitality  and  Governance  and  Leadership  exhibited  excellent  model  fit.  However,  these  tests  are  based  on  only  two  degrees  of  freedom  and  thus  may  be  underpowered  or  underspecified  to  provide  a  meaningful  test  of  the  model’s  validity.  The  four-­‐factor  model  provides  for  a  stronger,  more  powerful  test  of  the  theoretical  model.  The  four-­‐factor  CFA  model,  which  combined  the  subscale  models  into  a  single  analysis,  exhibited  statistical  misfit  but  borderline  approximate  model  fit.  However,  factor  inter-­‐correlations  were  notably  very  high  (.82  through  .93),  indicating  substantial  overlap  and  little  discriminant  validity  amongst  the  factors.  Therefore,  a  one-­‐factor  CFA  model  was  examined.  As  reflected  by  the  model  fit  indices,  the  one-­‐factor  model  fit  the  data  marginally  well.  Standardized  factor  loadings  and  associated  standard  errors  for  the  one-­‐factor  CFA  model  and  for  each  one-­‐factor  subscale  model  are  provided  in  Table  A.8.  Standardized  Factor  Loadings  for  the  1-­‐factor  CFA  Model  (Student-­‐level)  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9-­‐12  Grades)  in  Appendix  A:  Tables  from  the  Item-­‐level  Analyses  of  the  .  The  standardized  factor  loadings  were  uniformly  very  high,  with  the  vast  majority  greater  than  0.7  and  none  below  0.68.  These  numbers  indicate  that  each  item  is  a  salient  and  highly  reliable  measure  of  the  factor.  Moreover,  the  standardized  factor  loadings  are  estimated  with  a  moderate  degree  of  precision,  as  reflected  by  relatively  small  standard  errors.23      Though  the  CFA  analyses  and  higher  inter-­‐subscale  correlations  suggest  a  one-­‐factor  model,  it  is  possible  that  another  k-­‐factor  model  not  examined  generated  the  data.  Therefore,  an  exploratory  factor  analysis  (EFA)    was  performed  prior  to  settling  on  the  one-­‐factor  model  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­‐  12th  grades).  An  important  and  initial  task  in  conducting  exploratory  factor  analysis  is  to  determine  the  number  of  factors  or  dimensions  that  are  being  measured  by  the  survey  

                                                                                                               

22  The  SRMR  fit  index  is  not  available  for  these  analyses.  23  Given  evidence  of  potential  model  misspecification,  caution  in  interpreting  the  factor  loadings  is  warranted  until  these  results  can  be  cross-­‐validated  on  a  larger  data  set  and,  importantly,  conducted  at  the  school-­‐level.  

Page 31: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  31      

instrument.  This  determination  is  guided  by  statistical  tests  and  indices,  evaluation  of  eigenvalues,  and  meaningfulness  and  interpretability  of  the  solution.        Eigenvalues  quantify  the  variance  in  the  item  responses  that  is  explained  by  the  factors.  Factors  that  account  for  more  variation  are  considered  potentially  more  important  or  meaningful  than  factors  that  account  for  less  variation.  Eigenvalues  from  the  school-­‐level  factor  analysis  of  the  survey  data  are  reported  in  Table  12.  Eigenvalues  for  the  Exploratory  Factor  Analysis  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­‐  12th  grades).  An  often  cited  rule  of  thumb  is  the  Kaiser  rule,  which  states  that  factors  with  eigenvalues  greater  than  1.0  should  be  extracted.  According  to  the  Kaiser  rule,  a  model  with  six  factors  should  be  examined.  This  rule,  however,  has  been  criticized  as  leading  to  extraction  of  too  many  factors  and  thus  typically  should  be  considered  an  upper  bound  estimate  of  the  number  of  factors.      

Table  12.  Eigenvalues  for  the  Exploratory  Factor  Analysis  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­  12th  grades)  

    Factor   1   2   3   4   5   6   7   8   9   10   ...  

Eigenvalues       20.85   2.12   1.54   1.3   1.1   1.05   0.89   0.77   0.75   0.7    

 Another  factor  enumeration  approach  used  is  examination  of  the  Scree  plot,  depicted  in  Figure  3.  Scree  Plot  of  Eigenvalues  (School-­‐Level)  for  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­‐  12th  grades).  With  this  approach,  one  typically  seeks  the  point  where  there  is  a  pronounced  bend  (elbow)  in  the  curve.  Factors  before  the  bend  are  given  further  consideration.  According  to  the  scree  plot,  a  one-­‐factor  or  perhaps  a  two-­‐factor  model  should  be  considered  for  extraction.  Although  more  sophisticated  methods  for  factor  enumeration  exists  (e.g.,  parallel  analysis),  these  were  computationally  infeasible  or  inaccessible  for  the  present  analysis,  which  involved  ordinal  item  responses  distributions  that  make  it  difficult  to  employ  more  advanced  methods.  Thus,  the  one-­‐factor  and  two-­‐factor  models  were  given  closer  consideration.  

Page 32: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  32      

Figure  3.  Scree  Plot  of  Eigenvalues  (School-­Level)  for  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­  12th  grades)  

   In  conjunction  with  examination  of  the  eigenvalues  and  scree  plot,  statistical  tests  and  indices  of  model  fit  are  often  consulted.  All  models  exhibited  statistical  misfit  as  indicated  by  the  Chi-­‐square  test  of  exact  fit  (see  Table  13.  Model  Fit  Tests  and  Indices  for  EFA  Models  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­‐  12th  grades)).  With  most  indices  being  within  desired  ranges  for  all  models  under  consideration,  the  philosophical  principle  known  as  Occam's  razor  dictates  that  one  would  choose  the  most  parsimonious  model,  which  often  is  the  model  with  the  fewest  factors.  Consistent  with  the  CFA,  model  fit  indices,  in  conjunction  with  the  eigenvalues,  scree  plot  and  factor  pattern  matrix,  suggested  a  model  that  posits  a  single  factor  for  explaining  and  summarizing  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­‐  12th  grades)  item  responses.  This  conclusion  was  corroborated  by  the  less  interpretable  solutions  for  the  EFA  models  with  three  or  more  factors.  The  two  factor  solution  exhibited  a  factor  characterized  by  the  Catholic  Identity  and  Mission  items  and  item  32  from  the  Operational  Vitality  subscale  and  a  second  factor  characterized  by  the  remaining  items  and  item  10  (Catholic  Identity  and  Mission  subscale),  which  loaded  saliently  on  both  factors.  Although  relatively  interpretable,  the  factor  correlation  of  .74  was  considerably  large  and  suggests  that  there  may  not  be  two  conceptually  distinct  factors.  Therefore,  focus  is  on  the  one  factor  solution,  with  caution  given  the  possible  model  misspecification.  

Table  13.  Model  Fit  Tests  and  Indices  for  EFA  Models  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­  12th  grades)  

 1-­‐Factor  Model  

2-­‐Factor  Model  

3-­‐Factor  Model  

4-­‐Factor  Model  

5-­‐Factor  Model  

Chi-­‐Square   923.81   763.61   677.77   611.06   549.36  

df   560   526   493   461   430  

0  

5  

10  

15  

20  

25  

1   2   3   4   5   6   7   8   9   10  

Eigenvalue

 

Factor  

Page 33: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  33      

p-­‐value   <0.001   <0.001   <0.001   <0.001   <0.001  

RMSEA   0.080   0.067   0.061   0.056   0.052  

CFI   0.96   0.97   0.98   0.98   0.99  

SRMR   0.087   0.069   0.06   0.054   0.047  

Summary  statistics  for  the  distribution  of  composite  scores  formed  by  taking  the  average  of  item  scores  across  respondents  for  the  entire  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­‐  12th  grades)  and  each  subscale  can  be  found  in  Table  14.  Descriptive  Statistics  for  a  Total  Scale  Composite  Score  and  Subscale  Scores  (Student-­‐level)  Derived  from  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­‐  12th  grades).  The  theoretical  range  for  the  composite  and  subscale  scores  is  1  -­‐  5,  with  5  reflecting  greater  adherence  to  the  Standards  as  reported  by  students.  On  average,  students  tended  to  rate  their  schools  highly  on  the  total  scale  and  each  subscale.    

Table  14.  Descriptive  Statistics  for  a  Total  Scale  Composite  Score  and  Subscale  Scores  (Student-­level)  Derived  from  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­  12th  grades)  

 

Mean Standard Deviation Minimum Maximum 25th

Percentile Median 75th Percentile Reliability SEM

Mission  and  Catholic  Identity  

4.17   0.64   2.69   5.00   3.69   4.15   4.69   0.96   0.13  

Governance  and  Leadership  

4.18   0.79   1.00   5.00   4.00   4.25   4.75   0.93   0.21  

Academic  Excellence  

3.95   0.80   1.21   5.00   3.55   4.07   4.50   0.97   0.14  

Operational  Vitality  

3.79   0.94   1.25   5.00   3.44   4.00   4.50   0.91   0.28  

Total  Score   4.13   0.63   2.40   5.00   3.79   4.19   4.66   0.98   0.09  

The  composite  scale  and  subscale  scores  exhibited  excellent  reliability,  with  all  coefficients  estimated  >  .91.  This  is  largely  due  to  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­‐  12th  grades).  Based  on  an  estimate  of  the  standard  deviation  for  the  composite  (scale)  and  subscale  scores  and  their  respective  reliability  coefficients,  the  standard  errors  of  measurement  can  be  calculated  as

.  The  standard  error  of  measurement  can  be  used  to  form  confidence  bands  around  

scores  for  specific  schools.      

Observed  subscale  correlations  are  presented  in  Table  15.  Correlation  Matrix  for  Subscale  Scores  (Student-­‐level)  Derived  from  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­‐  12th  grades).  The  correlations  are  smaller  than  the  factor  inter-­‐correlations  reported  earlier  as  these  may  be  attenuated  by  measurement  error  and  are  obtained  using  a  different  estimator.  However,  these  correlations  were  also  very  high,  with  an  average  correlation  of  .85.  These  statistics  indicate  substantial  overlap  in  the  information  conveyed  by  the  subscale  scores  and  support  a  one-­‐factor  conceptualization  and  use  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­‐  12th  grades).    

Page 34: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  34      

Table  15.  Correlation  Matrix  for  Subscale  Scores  (Student-­level)  Derived  from  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­  12th  grades)  

 Mission  and  

Catholic  Identity  

Governance  and  

Leadership  

Academic  Excellence  

Operational  Vitality  

Mission  and  Catholic  Identity  

-­‐-­‐        

Governance  and  Leadership  

0.80   -­‐-­‐      

Academic  Excellence  

0.75   0.83   -­‐-­‐    

Operational  Vitality  

0.68   0.76   0.72   -­‐-­‐  

Note:  All  correlations  are  statistically  significant  at  p  <  .001.  

 

Summary  and  Recommendations  

Based  on  the  results  of  these  analyses,  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9th  -­‐  12th  grades)  appears  to  measure  a  single  construct  or  factor  reflecting  individual  perceptions  of  their  school’s  adherence  to  the  Standards.  All  items  were  supported  as  valid  and  reliable  indicators  of  a  general  Standards  construct  as  measured  at  the  respondent  level,  and  reliability  estimates  of  a  composite  score  computed  by  averaging  over  items  generally  indicated  high  reliability  (i.e.,  very  minimal  measurement  error).  In  fact,  the  attained  degree  of  reliability  affords  some  opportunity  to  shorten  the  scale  by  trimming  items  that  exhibit  high  rates  of  missing  data  (skipped  questions  and  Don’t  Know  responses)  to  ease  respondent  burden  while  maintaining  excellent  reliability.  Perhaps  contrary  to  initial  expectation,  there  was  no  evidence  that  this  survey  instrument  measures  four  distinct  factors  relating  to  the  four  domains  articulated  in  the  Standards  (viz.,  Mission  and  Catholic  Identity,  Governance  and  Leadership,  Academic  Excellence,  and  Operational  Vitality).  The  possible  explanations  and  implications  for  this  are  identical  to  those  described  for  the  other  Program  Effectiveness  surveys.  However,  limitations  of  the  9th  –  12th  grade  student  data  on  this  survey  limits  the  conclusions  that  can  be  drawn  from  these  results.  Too  few  high  schools  were  recruited  to  permit  analyses  at  the  school-­‐level,  the  desired  level  of  inference  for  these  surveys.  Analyses  were  thus  conducted  at  the  respondent  (i.e.,  student)  level,  but  it  is  very  possible  that  the  factor  structure  and  reliability  of  the  instrument  can  differ  at  the  school-­‐level  from  that  at  the  student-­‐level.  Moreover,  the  sample  size  was  modest  and  less  than  optimal  for  the  type  of  analyses  being  conducted.  As  such,  caution  is  warranted  in  generalizing  the  results  beyond  the  schools  and  respondents  sampled  to  the  broader  population  of  Catholic  high  schools.  It  is  recommended  that  a  multi-­‐level  factor  analysis  be  conducted  on  data  collected  from  a  larger  administration  of  this  survey  involving  at  least  100-­‐200  schools.      Certain  items,  particularly  those  from  the  Academic  Excellence  and  Operational  Vitality  subscales,  exhibited  high  rates  of  either  being  skipped  by  the  respondent  or  receiving  a  Don’t  Know  response.  Based  on  the  content  of  some  of  these  items  and  pattern  of  Don’t  Know  responses,  it  seems  likely  that  the  respondents  to  the  survey  may  not  have  adequate  knowledge  of  the  content  being  tapped  by  these  items  to  provide  a  rating.  To  the  extent  that  these  items  require  knowledge  of  their  school  that  the  respondents  may  not  have,  some  of  the  respondents  who  did  respond  to  the  item  may  have  done  so  with  little  actual  knowledge  

Page 35: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  35      

and  either  made  a  “guess”  or  answered  in  a  way  to  be  consistent  with  their  responses  on  other  items  or  overall  impression  of  their  school.  These  items  should  be  reviewed  and  considered  for  deletion  from  the  survey.24  Re-­‐writing  some  items  so  as  to  be  clearer  to  respondents  is  another  option.  It  would  be  desirable  to  have  missing  data  rates  less  than  10%.  The  worst  offending  items  with  the  rates  of  missing  data  greater  than  25%  are  listed  below:  

At  our  school,  teachers  work  together  to  use  data  on  student  performance  to  review  and  improve  courses  and  instruction.  

At  our  school,  teachers  work  together  to  help  each  other  become  better  teachers.   At  our  school,  all  administrators,  faculty  and  staff  engage  in  ongoing  professional  development.   At  our  school,  teachers  follow  school-­‐wide  policies  and  procedures  to  fairly  evaluate  and  

communicate  student  performance.   Our  school  communicates  how  well  students  are  achieving  in  comparison  to  similar  students  locally  

and/or  nationally.   Teachers  vary  the  types  of  assessments  they  use  to  monitor  individual  and  class-­‐wide  student  

learning.   Our  school  uses  different  ways  to  communicate  all  that  is  happening  in  our  school  to  

parents/guardians,  the  school  community  and  beyond.   Our  school  does  a  good  job  of  attracting  new  students  to  our  school.   Our  school  provides  programs  and  services  that  help  students  successfully  complete  the  school  

program  (for  example,  guidance  and  resource  programs).   Our  school  provides  opportunities  for  parents/guardians  to  be  involved  in  the  education  of  their  

students.  

The  above  items  were  associated  with  the  highest  rates  of  missing  data.  But  other  items  had  higher  than  desired  rates  of  missing  data  as  well.      A  5-­‐point  Likert  scale  was  employed  with  anchors  provided  for  only  the  extreme  points  (Strongly  Disagree,  Strongly  Agree).  Adding  labels  for  the  intermediate  points  may  (though  not  necessarily)  increase  the  range  of  the  scale  being  used  by  the  respondents,  which  may  translate  into  more  variability  in  ratings  and  subscale  scores  across  schools.  To  the  extent  that  the  increased  variability  is  reliable  variation,  adding  labels  for  all  points  would  tend  to  increase  reliability.  It  might  also  help  standardize  the  cognitive  mapping  used  by  respondents  to  link  scale  points  to  internal  sense  of  agreement  with  the  items.  This  may  increase  reliability  as  random  individual  differences  in  this  mapping  with  the  current  labeling  scheme  may  manifest  as  measurement  error.  Importantly,  using  standard  labels  (e.g.,  Strongly  Disagree,  Disagree,  Neither  Agree/Disagree,  Agree,  Strongly  Agree)  may  help  achieve  item  and  scale  scores  that  better  approximate  interval-­‐level  properties  and  do  so  more  consistently  so  across  respondents.  Ordinal  scales  only  convey  information  about  the  rank,  whereas  interval-­‐level  measures  also  convey  the  amount  of  difference.      

                                                                                                               

24  Note  that  deleting  reliable  items  can  reduce  subscale  reliability.  This  is  not  as  much  of  a  concern  for  scales  with  many  items,  but  may  be  for  scales  with  fewer  than  five  to  seven  items.  If  reliability  becomes  a  concern  after  deleting  items,  one  can  attempt  to  address  this  issue  by  generating  replacement  items  that  measure  the  construct  of  interest  for  those  that  are  deleted.  

Page 36: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  36      

 

Defining  Characteristics  Surveys  Description  

The  Defining  Characteristics  battery  of  surveys  is  designed  to  gauge  the  extent  to  which  Catholic  Schools  operate  in  accord  with  characteristics  articulated  by  the  Task  Force  for  National  Standards  and  Benchmarks  for  Effective  Catholic  Elementary  and  Secondary  Schools.25  Separate  self-­‐report  surveys  were  designed  to  capture  the  perceptions  of  adults,  middle  school  (5th  –  8th  grade)  students,  and  high  school  (9th  –  12th  grade)  students  of  their  schools’  “Catholic  identity”  with  respect  to  being  (a)  Centered  in  the  Person  of  Jesus  Christ,  (b)  Contributing  to  the  Evangelizing  Mission  of  the  Church,  (c)  Distinguished  by  Excellence,  (d)  Committed  to  Educate  the  Whole  Child,  (e)  Steeped  in  a  Catholic  Worldview,  (f)  Sustained  by  Gospel  Witness,  (g)  Shaped  by  Communion  and  Community,  (h)  Accessible  to  All  Students,  and  (j)  Established  by  the  Expressed  Authority  of  the  Bishop.  The  Catholic  School  –  Defining  Characteristics  surveys  each  consist  of  17  similar  or  identical  items.  Survey  respondents  self-­‐reported  their  perceptions  on  a  5-­‐point  Likert  response  scale  ranging  from  Strongly  Disagree  (1)  to  Strongly  Agree  (5).  Specific  items  and  their  subscale  grouping  can  be  found  in  Appendix  B:  Tables  from  the  Item-­‐level  Analyses  of  the  Catholic  School  –  Defining  Characteristics  Surveys.    

Adult  Survey  

Respondent  Characteristics  

A  total  of  2,080  respondents  from  52  schools  (average  of  40  respondents  per  school)  completed  the  Catholic  School  Defining  Characteristics  Survey  of  Adults  for  this  psychometric  study.  The  majority  of  respondents  were  non-­‐school  based  (74%).  Of  non-­‐school  based  respondents,  94.7%  were  parents,  3.1%  were  community  members,  and  1.6%  self-­‐identified  as  “other”.  Among  the  parent  respondents,  38.9%  reported  their  oldest  child’s  grade  as  being  4th  grade  or  below,  50.2%  in  5th  –  8th  grades,  and  10.9%  in  9th  –  12th  grades.  Of  school-­‐based  respondents  (26%  of  the  total  sample),  73.8%  were  teachers,  16.8%  were  staff,  and  9.4%  were  administrators.  Most  respondents  were  female  (84%),  and  the  racial  composition  of  the  sample  was:  White  (93.8%),  African  American  or  Black  (3.3%),  multi-­‐racial  (1.2%),  Asian  (1.2%),  American  Indian  or  Alaska  Native  (.3%),  or  Native  Hawaiian  or  Pacific  Islander  (.2%).  The  majority  of  respondents  identified  as  non-­‐Hispanic  (96.5%).    Most  respondents  expressed  their  religious  affiliation  as  Catholic  (86.9%).  In  general,  respondents  reported  multiple  years  of  affiliation  with  the  school:  29.2%  reported  1-­‐4  years,  31.9%  reported  5-­‐10  years,  and  29.3%  reported  more  than  10  years.  Only  9.6%  reported  less  than  1  year  of  affiliation  with  the  school.  

                                                                                                               

25  Center  for  Catholic  School  Effectiveness,  School  of  Education,  Loyola  University  Chicago,  in  partnership  with  the  Barbara  and  Patrick  Roche  Center  for  Catholic  Education,  School  of  Education,  Boston  College.  (2012).  National  Standards  and  Benchmarks  for  Effective  Catholic  Elementary  and  Secondary  Schools.  http://www.catholicschoolstandards.org/  

Page 37: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  37      

Response  Pattern  Summaries  

Response  frequencies  and  intraclass  correlation  coefficients  (ICCs)  for  each  of  the  17  items  are  reported  in  Table  B.3.  Item  Response  Frequencies  and  Intraclass  Correlation  Coefficients  (ICC)  for  the  Catholic  School  Defining  Characteristics  Survey  of  Adults  in  Appendix  B:  Tables  from  the  Item-­‐level  Analyses  of  the  Catholic  School  –  Defining  Characteristics  SurveysAppendix  .  On  all  items,  at  least  75%  of  adults  endorsed  the  Strongly  Agree  or  adjacent  response  options,  and  all  items  were  endorsed  (Strongly  Agree  or  adjacent  response)  by  at  least  77%  of  respondents.  Though  response  frequencies  are  clustered  towards  endorsement  of  options  indicating  greater  agreement,  there  was  some  degree  of  variability  of  frequency  patterns  across  items.  Item  percentages  for  skipped  or  Don't  Know  responses  ranged  between  3.9%  and  18.9%  (Mean  =  6.2%).  Two  items  were  skipped  or  had  a  Don't  Know  response  from  10%  or  more  of  respondents.  The  most  extreme  instances  were  item  16  (Our  school  does  everything  it  can  to  eliminate  obstacles  that  hinder  or  exclude  students  from  receiving  a  Catholic  education)  and  item  17  (Our  school  operates  with  the  expressed  approval  and  support  of  our  Bishop).  Of  note  is  that  the  rates  of  using  the  Don’t  Know  option  were  noticeably  higher  for  these  items,  particularly  item  17,  than  most  preceding  items,  which  may  indicate  that  respondents  did  not  feel  they  had  sufficient  knowledge  of  these  items  to  provide  their  ratings  on  these  items.  However  explained,  these  items  should  be  considered  for  rewording,  replacement  or  deletion  from  the  scale,  along  with  other  items  exhibiting  the  highest  skip  or  Don't  Know  rates,  given  the  problems  missing  data  presents  for  estimation  of  subscale  and  scale  scores.26    ICCs  quantify  the  proportion  of  variability  in  responses  that  is  attributable  to  variability  between  scores.  In  other  terms,  an  ICC  reflects  the  degree  of  non-­‐independence  amongst  responses  from  adults  affiliated  with  the  same  school,  with  0  =  independence  (i.e.,  no  systematic  variation  across  schools)  and  1  =  complete  dependence  (i.e.,  all  variation  in  responses  is  due  to  differences  across  schools).  As  this  survey  is  intended  to  measure  school-­‐level  alignment  with  the  Defining  Characteristics,  ICCs  greater  than  zero  are  to  be  expected  and  desired.  Additionally,  ICCs  greater  than  .01  support  the  need  for  statistical  methods  that  account  for  the  observed  non-­‐independence.  The  ICCs  for  most  items  were  moderate  in  magnitude,  reflecting  similarity  in  responses  from  adults  affiliated  the  same  school  and  justifying  the  need  for  statistical  methods  that  can  account  for  non-­‐independence  amongst  the  observations.  ICCs  across  all  items  ranged  from  .05  to  .13  (Mean  =  .08),  indicating  that  between  5%  and  13%  of  variation  in  item  responses  was  attributable  to  respondents  being  affiliated  with  different  schools.    The  Catholic  School  Defining  Characteristics  Survey  of  Adults  is  intended  to  provide  data  on  Catholic  schools’  alignment  to  the  Defining  Characteristics.  Given  this  intent,  the  focus  of  these  analyses  is  on  school-­‐level  aggregations  of  the  responses.  Table  B.4.  School-­‐level  Statistics  for  the  Catholic  School  Defining  Characteristics  Survey  of  Adults  in  Appendix  B:  Tables  from  the  Item-­‐level  Analyses  of  the  Catholic  School  –  Defining  Characteristics  Surveys  provides  statistics  to  describe  the  distributions  of  these  school-­‐level  aggregates,  where  adult  responses  for  each  item  are  averaged  with  other  adults  from  their  school.  The  theoretical  range  for  these  aggregated  item  averages  is  1  (Strongly  Disagree)  -­‐  5  (Strongly  Agree),  with  higher  scores  indicating  higher  within-­‐school  average  agreement  for  the  particular  item.  On  average,  

                                                                                                               

26  Although  most  analyses  described  in  the  Factorial  Validity  and  Reliability  use  statistical  procedures  that  can  incorporate  respondents  that  are  missing  data  on  one  or  more  items,  70  respondents  missing  data  on  all  items  were  necessarily  omitted  from  these  analyses.    

Page 38: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  38      

schools  were  rated  highly  on  all  of  the  items  (mean  average  rating  across  items  =  4.41)  and  minimum  ratings  averaged  3.30.  These  statistics  indicate  little  variability  in  mean  ratings  across  schools  for  the  items,  with  some  items  (#’s:  2,  4,  and  5)  exhibiting  very  little  variability.  Along  with  average,  minimum,  and  maximum  ratings,  standard  deviations  and  quartiles  are  also  provided  for  each  item.  The  median  (also  known  as  the  50th  percentile  or  second  quartile)  reflects  the  item  score  at  which  50%  of  schools  fall  at  or  below.  For  all  items,  50%  or  more  of  the  schools  scored  at  least  a  4.45.  Similarly  to  the  respondent-­‐level  item  statistics,  scores  were  clustered  toward  the  upper  end  of  the  distribution.  

Factorial  Validity  and  Reliability  

A  multilevel,  ordinal  confirmatory  factor  analyses  (CFA)  was  performed  on  the  Catholic  School  Defining  Characteristics  Survey  of  Adults.  A  single-­‐factor  model  of  the  survey  was  fit  to  the  data.27  This  model  fit  the  data  well  by  both  exact  fit  and  approximate  fit  standards  [χ2(119)  =  120.36,  p  =  .45,  RMSEA  =  0.002,  CFI  =  1.00,  SRMR  =  0.06]Multilevel,  ordinal  confirmatory  factor  analyses  (CFA)  began  with  the  separate  examination  of  Catholic  School  Program  Effectiveness  Survey  of  Adults  subscales  corresponding  to  each  of  the  four  Standards  for  Catholic  School  Program  Effectiveness.  These  CFA  models,  consisting  of  one-­‐factor  with  seven  to  fourteen  indicators  each,  were  fit  to  the  data.  Then,  these  models  were  combined  into  a  single,  four-­‐factor  CFA  model  of  the  Catholic  School  Program  Effectiveness  Survey  of  Adults.  CFA  concluded  with  estimation  of  a  model  where  all  items  were  specified  as  measures  of  a  single  factor  (one-­‐factor  CFA  model).  The  model  fit  tests  and  indices  for  these  analyses  are  reported  in  Table  1Error!  Not  a  valid  bookmark  self-­‐reference..  These  tests  evaluate  how  well  each  model  fits  the  data.  The  aim  is  to  identify  a  model  that  fits  the  data  well  (within  sampling  error)  or  approximately  well  and  is  parsimonious.  If  a  model  fits  the  data  poorly,  the  model  would  be  rejected  from  further  consideration.    Table  .  Still,  it  is  essential  that  the  one-­‐factor  model  solution  be  considered  meaningful  and  interpretable  where  the  meaningfulness  or  interpretability  of  a  solution  is  determined  by  considering  the  strength  and  pattern  of  relationships  between  the  items  and  underlying  (latent)  factor.  The  relationships  between  factors  and  indicators  are  typically  depicted  in  a  factor  loading  matrix.  Standardized  factor  loadings  and  associated  standard  errors  for  the  one-­‐factor  CFA  model  are  provided  in  Table  B.3.  Standardized  Factor  Loadings  for  the  1-­‐factor  CFA  Model  (School-­‐level)  of  the  Catholic  School  Defining  Characteristics  Survey  of  Adults  in  Appendix  B:  Tables  from  the  Item-­‐level  Analyses  of  the  Catholic  School  –  Defining  Characteristics  Surveys.  The  standardized  factor  loadings  were  uniformly  very  high,  with  most  loading  values  greater  than  0.8  and  none  below  0.66.  These  numbers  indicate  that  each  item  is  a  salient  and  highly  reliable  measure  of  the  factor.    Though  the  CFA  analysis  support  a  one-­‐factor  model,  it  is  possible  that  another  k-­‐factor  generated  the  data.  Therefore,  an  exploratory  factor  analysis  (EFA)  was  performed  prior  to  settling  on  the  one-­‐factor  model  of  the  Catholic  School  Defining  Characteristics  Survey  of  Adults.  An  important  and  initial  task  in  conducting  exploratory  factor  analysis  is  to  determine  the  number  of  factors  or  dimensions  that  are  being  measured  by  the  survey  instrument.  This  determination  is  guided  by  statistical  tests  and  indices,  evaluation  of  eigenvalues,  and  meaningfulness  and  interpretability  of  the  solution.  As  with  the  confirmatory  factor  analyses,  there  are  two-­‐levels  under  consideration,  respondent-­‐level  and  the  school-­‐level,  and  the  number  

                                                                                                               

27  In  these  analyses,  model  parameters  are  being  simultaneously  estimated  at  two-­‐levels,  respondent-­‐level  and  the  school-­‐level.,  with  an  unstructured  model  specified  for  the  within-­‐level  (i.e.,  no  over-­‐identifying  constraints  included).  However,  the  focus  of  these  analyses  is  on  discovering  the  factors  that  pertain  to  schools.    Thus,  the  focus  of  this  report  is  exclusively  on  results  pertaining  to  the  school-­‐level  models.  

Page 39: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  39      

of  factors  would  typically  be  determined  at  each  level.  However,  the  focus  of  this  report  is  on  the  school  as  the  unit  of  analysis.      Eigenvalues  quantify  the  variance  in  the  item  responses  that  is  explained  by  the  factors.  Factors  that  account  for  more  variation  are  considered  potentially  more  important  or  meaningful  than  factors  that  account  for  less  variation.  Eigenvalues  from  the  school-­‐level  factor  analysis  of  the  survey  data  are  reported  in  Table  15.  Eigenvalues  for  the  Exploratory  Factor  Analysis  of  the  Catholic  School  Defining  Characteristics  Survey  of  Adults.  An  often  cited  rule  of  thumb  is  the  Kaiser  rule,  which  states  that  factors  with  eigenvalues  greater  than  1.0  should  be  extracted.  According  to  the  Kaiser  rule,  a  model  with  two  factors  should  be  examined.  This  rule,  however,  has  been  criticized  as  leading  to  extraction  of  too  many  factors  and  thus  typically  should  be  considered  an  upper  bound  estimate  of  the  number  of  factors.      

Table  15.  Eigenvalues  for  the  Exploratory  Factor  Analysis  of  the  Catholic  School  Defining  Characteristics  Survey  of  Adults  

    Factor   1   2   3   4   5   6   7   8   9   10   ...  

Eigenvalues       13.55   1.04   0.7   0.52   0.46   0.36   0.23   0.14   0.13   0.1    

 Another  factor  enumeration  approach  used  is  examination  of  the  Scree  plot,  depicted  in  Figure  4.  Scree  Plot  of  Eigenvalues  (School-­‐Level)  for  the  Catholic  School  Defining  Characteristics  Survey  of  Adults.  With  this  approach,  one  typically  seeks  the  point  where  there  is  a  pronounced  bend  (elbow)  in  the  curve.  Factors  before  the  bend  are  given  further  consideration.  According  to  the  scree  plot,  a  one-­‐factor  or  perhaps  a  two-­‐factor  model  should  be  considered  for  extraction.  Although  more  sophisticated  methods  for  factor  enumeration  exists  (e.g.,  parallel  analysis),  these  were  computationally  infeasible  or  inaccessible  for  the  present  analysis,  which  involve  a  multilevel  structure  and  ordinal  item  responses  distributions  that  make  it  difficult  to  employ  more  advanced  methods.  Thus,  the  one-­‐factor  and  two-­‐factor  models  were  given  closer  consideration.  

Page 40: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  40      

Figure  4.  Scree  Plot  of  Eigenvalues  (School-­Level)  for  the  Catholic  School  Defining  Characteristics  Survey  of  Adults  

   In  conjunction  with  examination  of  the  eigenvalues  and  scree  plot,  statistical  tests  and  indices  of  model  fit  are  often  consulted.  All  models  fit  the  data  well  (see  Table  16.  Model  Fit  Tests  and  Indices  for  EFA  Models  of  the  Catholic  School  Defining  Characteristics  Survey  of  Adults).  With  most  indices  being  within  desired  ranges  for  all  models  under  consideration,  the  philosophical  principle  known  as  Occam's  razor  dictates  that  one  would  choose  the  most  parsimonious  model,  which  often  is  the  model  with  the  fewest  factors.  Consistent  with  the  CFA,  model  fit  indices,  in  conjunction  with  the  eigenvalues,  scree  plot  and  factor  pattern  matrix,  support  a  model  that  posits  a  single  factor  for  explaining  and  summarizing  the  Catholic  School  Program  Effectiveness  Survey  of  Adults  item  responses  when  aggregated  to  the  school-­‐level.  This  conclusion  was  corroborated  by  the  less  interpretable  solutions  for  the  EFA  models  with  two  factors  (e.g.,  factor  loading  >  1,  item  cross-­‐loadings,  factor  correlation  =  .82).  

Table  16.  Model  Fit  Tests  and  Indices  for  EFA  Models  of  the  Catholic  School  Defining  Characteristics  Survey  of  Adults  

 1-­‐Factor  Model  

2-­‐Factor  Model  

Chi-­‐Square   120.36   89.66  

df   119   103  

p-­‐value   0.45   0.82  

RMSEA   0.002   <0.001  

CFI   1.00   1.00  

SRMR   0.064   0.044  

0  

2  

4  

6  

8  

10  

12  

14  

16  

1   2   3   4   5   6   7   8   9   10  

Eigenvalue

 

Factor  

Page 41: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  41      

Summary  statistics  for  the  distribution  of  composite  score  formed  by  taking  the  average  of  school-­‐level  item  averages  for  the  entire  Catholic  School  Defining  Characteristics  Survey  of  Adults  can  be  found  in  Table  17.  Descriptive  Statistics  for  a  Total  Scale  Composite  Score  (School-­‐level)  Derived  from  the  Catholic  School  Defining  Characteristics  Survey  of  Adults.  The  theoretical  range  for  the  composite  and  subscale  scores  is  1  -­‐  5,  with  5  reflecting  greater  alignment  with  the  Defining  Characteristics  as  reported  by  the  respondents.  On  average,  schools  tended  to  score  highly  on  the  scale,  with  75%  or  more  schools  scoring  greater  than  4.3.    

Table  17.  Descriptive  Statistics  for  a  Total  Scale  Composite  Score  (School-­level)  Derived  from  the  Catholic  School  Defining  Characteristics  Survey  of  Adults  

 

Mean Standard Deviation Minimum Maximum 25th

Percentile Median 75th Percentile Reliability SEM

Total  Score   4.41   0.25   3.48   4.78   4.30   4.42   4.55   0.98   0.04  

The  composite  scale  exhibited  excellent  reliability  (.98),  largely  due  to  the  survey  having  a  large  number  of  highly  correlated  items,  each  being  rated  by  several  respondents  per  school.  Based  on  an  estimate  of  the  standard  deviation  for  the  composite  (scale)  and  subscale  scores  and  their  respective  reliability  coefficients,  

the  standard  errors  of  measurement  can  be  calculated  as .  The  standard  error  of  

measurement  can  be  used  to  form  confidence  bands  around  scores  for  specific  schools.    

Summary  and  Recommendations      Based  on  the  results  of  these  analyses,  the  Catholic  School  Defining  Characteristics  Survey  of  Adults  appears  to  measure  a  single  construct  or  factor  reflecting  the  shared  perceptions  of  schools’  alignment  with  the  Defining  Characteristics.  All  items  were  supported  as  valid  and  reliable  indicators  of  the  Defining  Characteristics  construct,  and  the  reliability  estimate  of  a  composite  score  computed  by  averaging  over  respondents  and  items  within  the  same  school  indicated  exceptionally  high  reliability  (i.e.,  very  minimal  measurement  error).  In  fact,  the  attained  degree  of  reliability  affords  some  opportunity  to  shorten  the  scale  by  trimming  items  that  exhibit  high  rates  of  missing  data  (skipped  questions  and  Don’t  Know  responses)  to  ease  respondent  burden  while  maintaining  excellent  reliability.      A  couple  items  exhibited  higher  than  desired  rates  of  either  being  skipped  by  the  respondent  or  receiving  a  Don’t  Know  response.  Based  on  the  content  of  the  items  and  pattern  of  Don’t  Know  responses,  it  seems  likely  that  the  respondents,  particularly  non-­‐school-­‐based  adults,  to  the  survey  may  not  have  adequate  knowledge  of  the  content  being  tapped  by  these  items.  To  the  extent  that  these  items  require  knowledge  of  their  school  that  the  respondents  may  not  have,  some  of  the  respondents  who  did  respond  to  the  item  may  have  done  so  with  little  actual  knowledge  and  either  made  a  “guess”  or  answered  in  a  way  to  be  consistent  with  their  responses  on  other  items  or  overall  impression  of  their  school.  These  items  should  be  reviewed  and  considered  for  deletion  from  the  survey.  It  would  be  desirable  to  have  missing  data  rates  less  than  10%.  The  items  with  rates  above  10%  are  listed  below:  

Our  school  does  everything  it  can  to  eliminate  obstacles  that  hinder  or  exclude  students  from  receiving  a  Catholic  education.  

Our  school  operates  with  the  expressed  approval  and  support  of  our  Bishop.  

Page 42: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  42      

A  5-­‐point  Likert  scale  was  employed  with  anchors  provided  for  only  the  extreme  points  (Strongly  Disagree,  Strongly  Agree).  Adding  labels  for  the  intermediate  points  may  (though  not  necessarily)  increase  the  range  of  the  scale  being  used  by  the  respondents,  which  may  translate  into  more  variability  in  ratings  and  subscale  scores  across  schools.  To  the  extent  that  the  increased  variability  is  reliable  variation,  adding  labels  for  all  points  would  tend  to  increase  reliability.  It  might  also  help  standardize  the  cognitive  mapping  used  by  respondents  to  link  scale  points  to  internal  sense  of  agreement  with  the  items.  This  may  increase  reliability  as  random  individual  differences  in  this  mapping  with  the  current  labeling  scheme  may  manifest  as  measurement  error.  Importantly,  using  standard  labels  (e.g.,  Strongly  Disagree,  Disagree,  Neither  Agree/Disagree,  Agree,  Strongly  Agree)  may  help  achieve  item  and  scale  scores  that  better  approximate  interval-­‐level  properties  and  do  so  more  consistently  so  across  respondents.  Ordinal  scales  only  convey  information  about  the  rank,  whereas  interval-­‐level  measures  also  convey  the  amount  of  difference.    Finally,  a  few  cautionary  statements  pertaining  to  these  analyses  are  necessary.  These  analyses  were  conducted  with  less  than  optimal  number  of  schools  for  assessing  the  school-­‐level  psychometric  properties  of  the  surveys.  Although  the  large  number  of  respondents  within  each  participating  school  and  the  generally  high  reliability  (and  factor  loadings)  of  most  items  facilitated  the  analysis,  it  would  have  been  desirable  to  have  at  least  150-­‐200  schools.  Related  and  also  impinging  on  the  results  is  that  the  high  non-­‐response  rate  of  schools  introduces  the  possibility  of  self-­‐selection  effects.  The  implication  of  this  is  that  the  obtained  sample  may  not  be  representative  of  the  broader  population  of  schools.  Therefore,  the  results  may  not  fully  generalize  to  that  broader  population.      

Page 43: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  43      

 Student  Survey  (Grades  5-­8)  

Respondent  Characteristics  

A  total  of  3,973  respondents  from  43  schools  (average  of  92  respondents  per  school)  completed  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (5th  -­‐  8th  grades)  for  this  psychometric  study.  Participating  students  were  nearly  evenly  distributed  across  grade  levels  (21.7%  to  27.5%  across  the  four  grades).  Just  over  half  of  respondents  were  female  (55%),  and  the  racial  composition  of  the  sample  was:  White  (81.7%),  African  American  or  Black  (5.5%),  multi-­‐racial  (7%),  Asian  (4.5%),  American  Indian  or  Alaska  Native  (.9%),  or  Native  Hawaiian  or  Pacific  Islander  (.4%).  The  majority  of  respondents  identified  as  non-­‐Hispanic  (92.4%).    Most  respondents  expressed  their  religious  affiliation  as  Catholic  (87.1%).  Most  respondents  also  reported  being  affiliated  with  the  school  for  more  than  4  years  (71.6%).  

Response  Pattern  Summaries  

Response  frequencies  and  intraclass  correlation  coefficients  (ICCs)  for  each  of  the  17  items  are  reported  in  Table  B.4.  Item  Response  Frequencies  and  Intraclass  Correlation  Coefficients  (ICC)  for  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (5-­‐8  Grades)  in  Appendix  B:  Tables  from  the  Item-­‐level  Analyses  of  the  Catholic  School  –  Defining  Characteristics  SurveysAppendix  .  On  13  of  the  items,  at  least  75%  of  middle  school  students  endorsed  the  Strongly  Agree  or  adjacent  response  options,  and  all  items  except  one  were  endorsed  (Strongly  Agree  or  adjacent  response)  by  at  least  50%  of  respondents.  Though  response  frequencies  are  clustered  towards  endorsement  of  options  indicating  greater  agreement,  there  was  some  degree  of  variability  of  frequency  patterns  across  items.  Item  percentages  for  skipped  or  Don't  Know  responses  ranged  between  2%  and  12.6%  (Mean  =  4.5%).  One  item  (viz.,  Our  school  operates  with  the  expressed  approval  and  support  of  our  Bishop)  had  both  the  highest  Don’t  Know  (9.4%)  and  the  highest  skip  (3.1%)  rates.  These  rates  were  much  higher  for  this  item  than  other  items,  indicating  that  respondents  did  not  feel  they  had  sufficient  knowledge  to  provide  their  rating  for  this  item.  This  item  should  be  considered  for  rewording,  replacement  or  deletion  from  the  scale.28    ICCs  quantify  the  proportion  of  variability  in  responses  that  is  attributable  to  variability  between  scores.  In  other  terms,  an  ICC  reflects  the  degree  of  non-­‐independence  amongst  responses  from  students  affiliated  with  the  same  school,  with  0  =  independence  (i.e.,  no  systematic  variation  across  schools)  and  1  =  complete  dependence  (i.e.,  all  variation  in  responses  is  due  to  differences  across  schools).  As  this  survey  is  intended  to  measure  school-­‐level  alignment  with  the  Defining  Characteristics,  ICCs  greater  than  zero  are  to  be  expected  and  desired.  Additionally,  ICCs  greater  than  .01  support  the  need  for  statistical  methods  that  account  for  the  observed  non-­‐independence.  The  ICCs  for  most  items  were  moderate  in  magnitude,  reflecting  similarity  in  responses  from  adults  affiliated  the  same  school  and  justifying  the  need  for  statistical  methods  that  can  account  for  non-­‐independence  amongst  the  observations.  ICCs  across  all  items  ranged  

                                                                                                               

28  Although  most  analyses  described  in  the  Factorial  Validity  and  Reliability  use  statistical  procedures  that  can  incorporate  respondents  that  are  missing  data  on  one  or  more  items,  62  respondents  missing  data  on  all  items  were  necessarily  omitted  from  these  analyses.    

Page 44: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  44      

from  .04  to  .18  (Mean  =  .08),  indicating  that  between  4%  and  18%  of  variation  in  item  responses  was  attributable  to  respondents  being  affiliated  with  different  schools.  

The  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­‐  8th  grades)  is  intended  to  provide  data  on  Catholic  schools’  adherence  to  the  Standards.  Given  this  intent,  the  focus  of  these  analyses  is  on  school-­‐level  aggregations  of  the  responses.  Table  B.5.  School-­‐level  Statistics  for  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (5-­‐8  Grades)  in  Appendix  B:  Tables  from  the  Item-­‐level  Analyses  of  the  Catholic  School  –  Defining  Characteristics  Surveys  provides  statistics  to  describe  the  distributions  of  these  school-­‐level  aggregates,  where  student  responses  for  each  item  are  averaged  with  other  students  from  their  school.  The  theoretical  range  for  these  aggregated  item  averages  is  1  (Strongly  Disagree)  -­‐  5  (Strongly  Agree),  with  higher  scores  indicating  higher  within-­‐school  average  agreement  for  the  particular  item.  On  average,  schools  were  rated  highly  on  all  of  the  items  (mean  average  rating  across  items  =  4.24)  and  minimum  ratings  averaged  2.91.  These  statistics  indicate  little  variability  in  mean  ratings  across  schools  for  the  items,  with  some  items  (#’s:  1,  4,  5,  and  9)  exhibiting  little  variability.  Along  with  average,  minimum,  and  maximum  ratings,  standard  deviations  and  quartiles  are  also  provided  for  each  item.  The  median  (also  known  as  the  50th  percentile  or  second  quartile)  reflects  the  item  score  at  which  50%  of  schools  fall  at  or  below.  For  all  items,  50%  or  more  of  the  schools  scored  at  least  a  3.31.  Similarly  to  the  respondent-­‐level  item  statistics,  scores  were  clustered  toward  the  upper  end  of  the  distribution.  

Factorial  Validity  and  Reliability  

A  multilevel,  ordinal  confirmatory  factor  analyses  (CFA)  was  performed  on  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (5th  -­‐  8th  grades).  A  single-­‐factor  model  of  the  survey  was  fit  to  the  data.29  This  model  fit  the  data  well  by  both  exact  fit  and  most  approximate  fit  standards  [χ2(119)  =  114.68,  p  =  .60,  RMSEA  <  0.001,  CFI  =  1.00,  SRMR  =  0.078]Multilevel,  ordinal  confirmatory  factor  analyses  (CFA)  began  with  the  separate  examination  of  Catholic  School  Program  Effectiveness  Survey  of  Adults  subscales  corresponding  to  each  of  the  four  Standards  for  Catholic  School  Program  Effectiveness.  These  CFA  models,  consisting  of  one-­‐factor  with  seven  to  fourteen  indicators  each,  were  fit  to  the  data.  Then,  these  models  were  combined  into  a  single,  four-­‐factor  CFA  model  of  the  Catholic  School  Program  Effectiveness  Survey  of  Adults.  CFA  concluded  with  estimation  of  a  model  where  all  items  were  specified  as  measures  of  a  single  factor  (one-­‐factor  CFA  model).  The  model  fit  tests  and  indices  for  these  analyses  are  reported  in  Table  1Error!  Not  a  valid  bookmark  self-­‐reference..  These  tests  evaluate  how  well  each  model  fits  the  data.  The  aim  is  to  identify  a  model  that  fits  the  data  well  (within  sampling  error)  or  approximately  well  and  is  parsimonious.  If  a  model  fits  the  data  poorly,  the  model  would  be  rejected  from  further  consideration.    Table  .  Still,  it  is  essential  that  the  one-­‐factor  model  solution  be  considered  meaningful  and  interpretable  where  the  meaningfulness  or  interpretability  of  a  solution  is  determined  by  considering  the  strength  and  pattern  of  relationships  between  the  items  and  underlying  (latent)  factor.  The  relationships  between  factors  and  indicators  are  typically  depicted  in  a  factor  loading  matrix.  Standardized  factor  loadings  and  associated  standard  errors  for  the  one-­‐factor  CFA  model  are  provided  in  Table  in  Appendix  B:  Tables  from  the  Item-­‐level  Analyses  of  the  Catholic  School  –  Defining  Characteristics  Surveys.  The  standardized  factor  

                                                                                                               

29  In  these  analyses,  model  parameters  are  being  simultaneously  estimated  at  two-­‐levels,  respondent-­‐level  and  the  school-­‐level.,  with  an  unstructured  model  specified  for  the  within-­‐level  (i.e.,  no  over-­‐identifying  constraints  included).  However,  the  focus  of  these  analyses  is  on  discovering  the  factors  that  pertain  to  schools.    Thus,  the  focus  of  this  report  is  exclusively  on  results  pertaining  to  the  school-­‐level  models.  

Page 45: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  45      

loadings  were  uniformly  very  high,  with  most  loading  values  greater  than  0.8  and  none  below  0.47.  These  numbers  indicate  that  each  item  is  a  salient  and  highly  reliable  measure  of  the  factor.      Though  the  CFA  analysis  support  a  one-­‐factor  model,  it  is  possible  that  another  k-­‐factor  generated  the  data.  Therefore,  an  exploratory  factor  analysis  (EFA)  was  performed  prior  to  settling  on  the  one-­‐factor  model  of  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (5th  -­‐  8th  grades).  An  important  and  initial  task  in  conducting  exploratory  factor  analysis  is  to  determine  the  number  of  factors  or  dimensions  that  are  being  measured  by  the  survey  instrument.  This  determination  is  guided  by  statistical  tests  and  indices,  evaluation  of  eigenvalues,  and  meaningfulness  and  interpretability  of  the  solution.  As  with  the  confirmatory  factor  analyses,  there  are  two-­‐levels  under  consideration,  respondent-­‐level  and  the  school-­‐level,  and  the  number  of  factors  would  typically  be  determined  at  each  level.  However,  the  focus  of  this  report  is  on  the  school  as  the  unit  of  analysis.        Eigenvalues  quantify  the  variance  in  the  item  responses  that  is  explained  by  the  factors.  Factors  that  account  for  more  variation  are  considered  potentially  more  important  or  meaningful  than  factors  that  account  for  less  variation.  Eigenvalues  from  the  school-­‐level  factor  analysis  of  the  survey  data  are  reported  in  Table  18.  Eigenvalues  for  the  Exploratory  Factor  Analysis  of  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (5th  –  8th  grades).  An  often  cited  rule  of  thumb  is  the  Kaiser  rule,  which  states  that  factors  with  eigenvalues  greater  than  1.0  should  be  extracted.  According  to  the  Kaiser  rule,  a  model  with  three  factors  should  be  examined.  This  rule,  however,  has  been  criticized  as  leading  to  extraction  of  too  many  factors  and  thus  typically  should  be  considered  an  upper  bound  estimate  of  the  number  of  factors.      

Table  18.  Eigenvalues  for  the  Exploratory  Factor  Analysis  of  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (5th  –  8th  grades)  

    Factor   1   2   3   4   5   6   7   8   9   10   ...  

Eigenvalues       11.86   1.23   1.07   0.86   0.53   0.51   0.41   0.22   0.19   0.18    

Another  factor  enumeration  approach  used  is  examination  of  the  Scree  plot,  depicted  in  Figure  5.  Scree  Plot  of  Eigenvalues  (School-­‐Level)  for  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (5th  –  8th  grades).  With  this  approach,  one  typically  seeks  the  point  where  there  is  a  pronounced  bend  (elbow)  in  the  curve.  Factors  before  the  bend  are  given  further  consideration.  According  to  the  scree  plot,  a  one-­‐factor  or  perhaps  a  two-­‐factor  model  should  be  considered  for  extraction.  Although  more  sophisticated  methods  for  factor  enumeration  exists  (e.g.,  parallel  analysis),  these  were  computationally  infeasible  or  inaccessible  for  the  present  analysis,  which  involve  a  multilevel  structure  and  ordinal  item  responses  distributions.  Thus,  the  one-­‐  and  two-­‐factor  models  were  given  closer  consideration.  

Page 46: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  46      

Figure  5.  Scree  Plot  of  Eigenvalues  (School-­Level)  for  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (5th  –  8th  grades)  

   In  conjunction  with  examination  of  the  eigenvalues  and  scree  plot,  statistical  tests  and  indices  of  model  fit  are  often  consulted.  All  models  fit  the  data  well  (see  Table  19.  Model  Fit  Tests  and  Indices  for  EFA  Models  of  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (5th  –  8th  grades)).  With  most  indices  being  within  desired  ranges  for  all  models  under  consideration,  the  philosophical  principle  known  as  Occam's  razor  dictates  that  one  would  choose  the  most  parsimonious  model,  which  often  is  the  model  with  the  fewest  factors.  Consistent  with  the  CFA,  model  fit  indices,  in  conjunction  with  the  eigenvalues,  scree  plot  and  factor  pattern  matrix,  support  a  model  that  posits  a  single  factor  for  explaining  and  summarizing  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5th  -­‐  8th  grades)  item  responses  when  aggregated  to  the  school-­‐level.  This  conclusion  was  corroborated  by  the  less  interpretable  solutions  for  the  EFA  model  with  two  factors  (e.g.,  factor  loading  >  1,  item  cross-­‐loadings,  high  factor  correlation  =  .65).  

Table  19.  Model  Fit  Tests  and  Indices  for  EFA  Models  of  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (5th  –  8th  grades)  

 1-­‐Factor  Model  

2-­‐Factor  Model  

Chi-­‐Square   114.68   90.87  

df   119   103  

p-­‐value   0.60   0.80  

RMSEA   <  .0001   <.0001  

CFI   1.00   1.00  

SRMR   0.078   0.062  

0  

2  

4  

6  

8  

10  

12  

14  

1   2   3   4   5   6   7   8   9   10  

Eigenvalue

 

Factor  

Page 47: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  47      

Summary  statistics  for  the  distribution  of  composite  score  formed  by  taking  the  average  of  school-­‐level  item  averages  for  the  entire  Catholic  School  Defining  Characteristics  Survey  of  Students  (5th  -­‐  8th  grades)  can  be  found  in  Table  20.  Descriptive  Statistics  for  a  Total  Scale  Composite  Score  (School-­‐level)  Derived  from  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (5th  –  8th  grades).  The  theoretical  range  for  the  composite  scores  is  1  -­‐  5,  with  5  reflecting  greater  alignment  with  the  Defining  Characteristics  as  reported  by  students.  On  average,  schools  tended  to  score  highly  on  the  scale,  with  75%  or  more  schools  scoring  greater  than  4.12.    

Table  20.  Descriptive  Statistics  for  a  Total  Scale  Composite  Score  (School-­level)  Derived  from  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (5th  –  8th  grades)  

 

Mean Standard Deviation Minimum Maximum 25th

Percentile Median 75th Percentile Reliability SEM

Total  Score   4.24   0.30   3.06   4.77   4.12   4.25   4.42   0.97   0.05  

 The  composite  scale  exhibited  excellent  reliability  (.97),  largely  due  to  the  survey  having  a  large  number  of  highly  correlated  items,  each  being  rated  by  several  respondents  per  school.  Based  on  an  estimate  of  the  standard  deviation  for  the  composite  (scale)  and  subscale  scores  and  their  respective  reliability  coefficients,  

the  standard  errors  of  measurement  can  be  calculated  as .  The  standard  error  of  

measurement  can  be  used  to  form  confidence  bands  around  scores  for  specific  schools.  

Summary  and  Recommendations  

Based  on  the  results  of  these  analyses,  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (5th  -­‐  8th  grades)  appears  to  measure  a  single  construct  or  factor  reflecting  the  shared  perceptions  of  schools’  alignment  with  the  Defining  Characteristics.  All  items  were  supported  as  valid  and  reliable  indicators  of  the  Defining  Characteristics  construct,  and  the  reliability  estimate  of  a  composite  score  computed  by  averaging  over  respondents  and  items  within  the  same  school  indicated  exceptionally  high  reliability  (i.e.,  very  minimal  measurement  error).  In  fact,  the  attained  degree  of  reliability  affords  some  opportunity  to  shorten  the  scale  by  trimming  items  that  exhibit  high  rates  of  missing  data  (skipped  questions  and  Don’t  Know  responses)  to  ease  respondent  burden  while  maintaining  excellent  reliability.      One  item  (viz.,  Our  school  operates  with  the  expressed  approval  and  support  of  our  Bishop)  exhibited  a  higher  than  desired  rate  of  either  being  skipped  by  the  respondent  or  receiving  a  Don’t  Know  response.  It  seems  likely  that  the  respondents  to  the  survey  may  not  have  adequate  knowledge  of  the  content  being  tapped  by  this  item.  To  the  extent  that  this  items  require  knowledge  of  their  school  that  the  respondents  may  not  have,  some  of  the  respondents  who  did  respond  to  the  item  may  have  done  so  with  little  actual  knowledge  and  either  made  a  “guess”  or  answered  in  a  way  to  be  consistent  with  their  responses  on  other  items  or  overall  impression  of  their  school.  This  item  should  be  reviewed  and  considered  for  deletion  from  the  survey.  It  would  be  desirable  to  have  missing  data  rates  less  than  10%.      A  5-­‐point  Likert  scale  was  employed  with  anchors  provided  for  only  the  extreme  points  (Strongly  Disagree,  Strongly  Agree).  Adding  labels  for  the  intermediate  points  may  (though  not  necessarily)  increase  the  range  of  the  scale  being  used  by  the  respondents,  which  may  translate  into  more  variability  in  ratings  and  subscale  scores  across  schools.  To  the  extent  that  the  increased  variability  is  reliable  variation,  adding  

Page 48: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  48      

labels  for  all  points  would  tend  to  increase  reliability.  It  might  also  help  standardize  the  cognitive  mapping  used  by  respondents  to  link  scale  points  to  internal  sense  of  agreement  with  the  items.  This  may  increase  reliability  as  random  individual  differences  in  this  mapping  with  the  current  labeling  scheme  may  manifest  as  measurement  error.  Importantly,  using  standard  labels  (e.g.,  Strongly  Disagree,  Disagree,  Neither  Agree/Disagree,  Agree,  Strongly  Agree)  may  help  achieve  item  and  scale  scores  that  better  approximate  interval-­‐level  properties  and  do  so  more  consistently  so  across  respondents.  Ordinal  scales  only  convey  information  about  the  rank,  whereas  interval-­‐level  measures  also  convey  the  amount  of  difference.    Finally,  a  few  cautionary  statements  pertaining  to  these  analyses  are  necessary.  These  analyses  were  conducted  with  less  than  optimal  number  of  schools  for  assessing  the  school-­‐level  psychometric  properties  of  the  surveys.  Although  the  large  number  of  respondents  within  each  participating  school  and  the  generally  high  reliability  (and  factor  loadings)  of  most  items  facilitated  the  analysis,  it  would  have  been  desirable  to  have  at  least  150-­‐200  schools.  Related  and  also  impinging  on  the  results  is  that  the  high  non-­‐response  rate  of  schools  introduces  the  possibility  of  self-­‐selection  effects.  The  implication  of  this  is  that  the  obtained  sample  may  not  be  representative  of  the  broader  population  of  schools.  Therefore,  the  results  may  not  fully  generalize  to  that  broader  population.  

 

Page 49: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  49      

 Student  Survey  (Grades  9-­12)  

Respondent  Characteristics  

A  total  of  128  respondents  from  5  schools  (average  of  26  respondents  per  school)  completed  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (9th  -­‐  12th  grades)  for  this  psychometric  study.  Participating  students  were  mostly  in  the  9th  (42%)  or  11th  (25%)  grades.  The  majority  of  participants  were  female  (66.4%),  and  the  racial  composition  of  the  sample  was:  White  (82%),  African  American  or  Black  (9.4%),  multi-­‐racial  (5.5%),  or  Asian  (3.1%).  The  majority  of  respondents  identified  as  non-­‐Hispanic  (94.5%).    Most  respondents  expressed  their  religious  affiliation  as  Catholic  (77.3%).  Respondents  were  fairly  uniformly  distributed  in  their  number  of  years  attending  the  school  (10.9%  -­‐  21.1%).    

Response  Pattern  Summaries  

Response  frequencies  for  each  of  the  17  items  are  reported  in  Table  B.7.  Item  Response  Frequencies  for  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (9-­‐12  Grades)  in  Appendix  B:  Tables  from  the  Item-­‐level  Analyses  of  the  Catholic  School  –  Defining  Characteristics  SurveysAppendix  .  On  all  items,  at  least  75%  of  high  school  students  endorsed  the  Strongly  Agree  or  adjacent  response  options.  Though  response  frequencies  are  clustered  towards  endorsement  of  options  indicating  greater  agreement,  there  was  some  degree  of  variability  of  frequency  patterns  across  items.  Item  percentages  for  skipped  or  Don't  Know  responses  ranged  between  5.5%  and  12.5%  (Mean  =  7.0%).  The  items  with  more  than  10%  missing  data  were:  (a)  The  adults  in  our  school  show  that  they  are  partners  with  our  parents/guardians  in  our  Catholic  education,  (b)  Our  school  does  everything  it  can  to  eliminate  obstacles  that  hinder  or  exclude  students  from  receiving  a  Catholic  education,  and  (c)  Our  school  operates  with  the  expressed  approval  and  support  of  our  Bishop.  These  items  may  ask  respondents  about  aspects  which  they  have  little  or  no  knowledge  on  which  to  base  a  rating.  These  items  were  among  a  few  items  with  a  non-­‐zero  rate  of  Don’t  Know  response.  These  items  should  be  considered  for  rewording,  replacement  or  deletion  from  the  scale.30    

The  Catholic  School  Defining  Characteristics  Survey  of  Students  (9th  -­‐  12th  grades)  is  intended  to  provide  data  on  Catholic  schools’  alignment  to  the  guiding  Defining  Characteristics  of  Catholic  schools.  Given  this  intent,  the  desired  focus  of  analysis  is  on  school-­‐level  aggregations  of  the  responses.  However,  too  few  high  schools  were  successfully  recruited  for  this  psychometric  study  to  permit  school-­‐level  analyses.  Therefore,  all  subsequent  analyses  report  on  analyses  at  the  respondent-­‐level.  In  interpreting  the  results  of  these  analyses,  understand  that  these  results  may  not  extend  to  the  school-­‐level  and  any  such  generalization  is  discouraged.31  

                                                                                                               

30  Although  most  analyses  described  in  the  Factorial  Validity  and  Reliability  use  statistical  procedures  that  can  incorporate  respondents  that  are  missing  data  on  one  or  more  items,  seven  respondents  missing  data  on  all  items  were  necessarily  omitted  from  these  analyses.  31  In  addition,  the  respondent-­‐level  sample  size  was  particularly  modest  for  these  analyses.  Caution  is  warranted  when  interpreting  the  results  until  more  data  is  available.  

Page 50: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  50      

Factorial  Validity  and  Reliability  

An  ordinal  confirmatory  factor  analyses  (CFA)  was  performed  on  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (9th  -­‐  12th  grades).  A  single-­‐factor  model  of  the  survey  was  fit  to  the  data.  The  model  exhibited  statistically  significant  misfit  and  mixed  results  on  the  approximate  fit  indices  [χ2(119)  =  237.01,  p  <  .001,  RMSEA  =  0.091,  CFI  =  0.97,  SRMR  =  0.057]Multilevel,  ordinal  confirmatory  factor  analyses  (CFA)  began  with  the  separate  examination  of  Catholic  School  Program  Effectiveness  Survey  of  Adults  subscales  corresponding  to  each  of  the  four  Standards  for  Catholic  School  Program  Effectiveness.  These  CFA  models,  consisting  of  one-­‐factor  with  seven  to  fourteen  indicators  each,  were  fit  to  the  data.  Then,  these  models  were  combined  into  a  single,  four-­‐factor  CFA  model  of  the  Catholic  School  Program  Effectiveness  Survey  of  Adults.  CFA  concluded  with  estimation  of  a  model  where  all  items  were  specified  as  measures  of  a  single  factor  (one-­‐factor  CFA  model).  The  model  fit  tests  and  indices  for  these  analyses  are  reported  in  Table  1Error!  Not  a  valid  bookmark  self-­‐reference..  These  tests  evaluate  how  well  each  model  fits  the  data.  The  aim  is  to  identify  a  model  that  fits  the  data  well  (within  sampling  error)  or  approximately  well  and  is  parsimonious.  If  a  model  fits  the  data  poorly,  the  model  would  be  rejected  from  further  consideration.    Table  .  Modification  indices  and  residuals  were  examined,  but  these  were  not  informative  as  to  source  of  misfit.  Standardized  factor  loadings  and  associated  standard  errors  for  this  model  are  provided  in  Table    in  Appendix  B:  Tables  from  the  Item-­‐level  Analyses  of  the  Catholic  School  –  Defining  Characteristics  Surveys,  but  they  should  be  interpreted  with  caution  given  the  evidence  of  possible  misfit.  The  standardized  factor  loadings  were  uniformly  very  high,  with  most  loading  values  greater  than  0.8  and  none  below  0.75.      It  is  possible  that  another  k-­‐factor  generated  the  data.  Therefore,  an  exploratory  factor  analysis  (EFA)  was  performed.  An  important  and  initial  task  in  conducting  exploratory  factor  analysis  is  to  determine  the  number  of  factors  or  dimensions  that  are  being  measured  by  the  survey  instrument.  This  determination  is  guided  by  statistical  tests  and  indices,  evaluation  of  eigenvalues,  and  meaningfulness  and  interpretability  of  the  solution.  As  with  the  confirmatory  factor  analyses,  there  are  two-­‐levels  under  consideration,  respondent-­‐level  and  the  school-­‐level,  and  the  number  of  factors  would  typically  be  determined  at  each  level.  However,  the  focus  of  this  report  is  on  the  school  as  the  unit  of  analysis.        Eigenvalues  quantify  the  variance  in  the  item  responses  that  is  explained  by  the  factors.  Factors  that  account  for  more  variation  are  considered  potentially  more  important  or  meaningful  than  factors  that  account  for  less  variation.  Eigenvalues  from  the  school-­‐level  factor  analysis  of  the  survey  data  are  reported  in  Table  21.  Eigenvalues  for  the  Exploratory  Factor  Analysis  of  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (9th  –  12th  grades).  An  often  cited  rule  of  thumb  is  the  Kaiser  rule,  which  states  that  factors  with  eigenvalues  greater  than  1.0  should  be  extracted.  According  to  the  Kaiser  rule,  a  model  with  one  factor  should  be  examined  further.    

Table  21.  Eigenvalues  for  the  Exploratory  Factor  Analysis  of  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (9th  –  12th  grades)  

    Factor   1   2   3   4   5   6   7   8   9   10   ...  

Eigenvalues       12.36   0.94   0.65   0.49   0.46   0.43   0.37   0.28   0.26   0.2   …  

 Another  factor  enumeration  approach  used  is  examination  of  the  Scree  plot,  depicted  in  Figure  6.  Scree  Plot  of  Eigenvalues  (School-­‐Level)  for  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (9th  –  

Page 51: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  51      

12th  grades).  With  this  approach,  one  typically  seeks  the  point  where  there  is  a  pronounced  bend  (elbow)  in  the  curve.  Factors  before  the  bend  are  given  further  consideration.  According  to  the  scree  plot,  a  one-­‐factor  model  should  be  considered.  Although  more  sophisticated  methods  for  factor  enumeration  exists  (e.g.,  parallel  analysis),  these  were  computationally  infeasible  or  inaccessible  for  the  present  analysis,  which  involve  ordinal  item  responses  distributions  that  make  it  difficult  to  employ  more  advanced  methods.  This  model  is  identical  to  the  one-­‐factor  CFA  model.  

Figure  6.  Scree  Plot  of  Eigenvalues  (School-­Level)  for  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (9th  –  12th  grades)  

   For  comparison,  the  fit  indices  for  the  one-­‐  and  two-­‐factor  EFA  models  are  provided  in  Table  22.  Model  Fit  Tests  and  Indices  for  EFA  Models  of  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (9th  –  12th  grades).  Although  the  one-­‐factor  model  exhibits  evidence  of  misfit,  the  EFA  results  provide  more  support  for  this  model  over  another  k-­‐class  model.    

Table  22.  Model  Fit  Tests  and  Indices  for  EFA  Models  of  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (9th  –  12th  grades)  

 1-­‐Factor  Model  

2-­‐Factor  Model  

Chi-­‐Square   237.01   162.01  

df   119   103  

p-­‐value   <  .0001   0.0002  

RMSEA   0.091   0.069  

CFI   0.98   0.99  

SRMR   0.057   0.040  

0  

2  

4  

6  

8  

10  

12  

14  

1   2   3   4   5   6   7   8   9   10  

Eigenvalue

 

Factor  

Page 52: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  52      

Summary  statistics  for  the  distribution  of  composite  score  formed  by  taking  the  average  of  student-­‐level  item  averages  for  the  entire  Catholic  School  Defining  Characteristics  Survey  of  Students  (9th  -­‐  12th  grades)  can  be  found  in  Table  23.  Descriptive  Statistics  for  a  Total  Scale  Composite  Score  (Student-­‐level)  Derived  from  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (9th  –  12th  grades).  The  theoretical  range  for  the  composite  is  1  -­‐  5,  with  5  reflecting  greater  alignment  with  the  Defining  Characteristics  as  reported  by  students.  On  average,  students  tended  to  rate  their  schools  highly  on  the  scale,  with  75%  or  more  students  rating  their  school  higher  than  4.00.    

Table  23.  Descriptive  Statistics  for  a  Total  Scale  Composite  Score  (Student-­level)  Derived  from  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (9th  –  12th  grades)  

 

Mean Standard Deviation Minimum Maximum 25th

Percentile Median 75th Percentile Reliability SEM

Total  Score   4.41   0.65   1.00   5.00   4.00   4.59   4.99   0.98   0.09  

 The  composite  scale  exhibited  excellent  reliability  (.98),  largely  due  to  the  survey  having  a  large  number  of  highly  correlated  items,  each  being  rated  by  several  respondents  per  school.  Based  on  an  estimate  of  the  standard  deviation  for  the  composite  (scale)  and  subscale  scores  and  their  respective  reliability  coefficients,  

the  standard  errors  of  measurement  can  be  calculated  as .  The  standard  error  of  

measurement  can  be  used  to  form  confidence  bands  around  scores  for  specific  schools.    

Summary  and  Recommendations  

Based  on  the  results  of  these  analyses,  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (9th  -­‐  12th  grades)  appears  to  measure  a  single  construct  or  factor  reflecting  the  shared  perceptions  of  schools’  alignment  with  the  Defining  Characteristics,  through  model  fit  tests  and  statistics  were  not  conclusive.  All  items  were  supported  as  valid  and  reliable  indicators  of  the  Defining  Characteristics  construct  as  measured  at  the  respondent  level,  and  the  reliability  estimate  of  a  composite  score  computed  by  averaging  over  items  generally  indicated  high  reliability  (i.e.,  very  minimal  measurement  error).  In  fact,  the  attained  degree  of  reliability  affords  some  opportunity  to  shorten  the  scale  by  trimming  items  that  exhibit  high  rates  of  missing  data  (skipped  questions  and  Don’t  Know  responses)  to  ease  respondent  burden  while  maintaining  excellent  reliability.  However,  limitations  of  the  9th  –  12th  grade  student  data  on  this  survey  limits  the  conclusions  that  can  be  drawn  from  these  results.  Too  few  high  schools  were  recruited  to  permit  analyses  at  the  school-­‐level,  the  desired  level  of  inference  for  these  surveys.  Analyses  were  thus  conducted  at  the  respondent  (i.e.,  student)  level,  but  it  is  very  possible  that  the  factor  structure  and  reliability  of  the  instrument  can  differ  at  the  school-­‐level  from  that  at  the  student-­‐level.  Moreover,  the  sample  size  was  modest  and  less  than  optimal  for  the  type  of  analyses  being  conducted.  As  such,  caution  is  warranted  in  generalizing  the  results  beyond  the  schools  and  respondents  sampled  to  the  broader  population  of  Catholic  high  schools.  It  is  recommended  that  a  multi-­‐level  factor  analysis  be  conducted  on  data  collected  from  a  larger  administration  of  this  survey  involving  at  least  100-­‐200  schools.      A  few  items  exhibited  higher  than  desired  rates  of  either  being  skipped  by  the  respondent  or  receiving  a  Don’t  Know  response.  Based  on  the  content  of  the  items  and  pattern  of  Don’t  Know  responses,  it  seems  likely  that  the  respondents,  particularly  non-­‐school-­‐based  adults,  to  the  survey  may  not  have  adequate  

Page 53: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  53      

knowledge  of  the  content  being  tapped  by  these  items.  To  the  extent  that  these  items  require  knowledge  of  their  school  that  the  respondents  may  not  have,  some  of  the  respondents  who  did  respond  to  the  item  may  have  done  so  with  little  actual  knowledge  and  either  made  a  “guess”  or  answered  in  a  way  to  be  consistent  with  their  responses  on  other  items  or  overall  impression  of  their  school.  These  items  should  be  reviewed  and  considered  for  deletion  from  the  survey.  It  would  be  desirable  to  have  missing  data  rates  less  than  10%.  The  items  with  rates  above  10%  are  listed  below:  

The  adults  in  our  school  show  that  they  are  partners  with  our  parents/guardians  in  our  Catholic  education.  

Our  school  does  everything  it  can  to  eliminate  obstacles  that  hinder  or  exclude  students  from  receiving  a  Catholic  education.  

Our  school  operates  with  the  expressed  approval  and  support  of  our  Bishop.  

A  5-­‐point  Likert  scale  was  employed  with  anchors  provided  for  only  the  extreme  points  (Strongly  Disagree,  Strongly  Agree).  Adding  labels  for  the  intermediate  points  may  (though  not  necessarily)  increase  the  range  of  the  scale  being  used  by  the  respondents,  which  may  translate  into  more  variability  in  ratings  and  subscale  scores  across  schools.  To  the  extent  that  the  increased  variability  is  reliable  variation,  adding  labels  for  all  points  would  tend  to  increase  reliability.  It  might  also  help  standardize  the  cognitive  mapping  used  by  respondents  to  link  scale  points  to  internal  sense  of  agreement  with  the  items.  This  may  increase  reliability  as  random  individual  differences  in  this  mapping  with  the  current  labeling  scheme  may  manifest  as  measurement  error.  Importantly,  using  standard  labels  (e.g.,  Strongly  Disagree,  Disagree,  Neither  Agree/Disagree,  Agree,  Strongly  Agree)  may  help  achieve  item  and  scale  scores  that  better  approximate  interval-­‐level  properties  and  do  so  more  consistently  so  across  respondents.  Ordinal  scales  only  convey  information  about  the  rank,  whereas  interval-­‐level  measures  also  convey  the  amount  of  difference.    

Page 54: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  54      

 

Appendix  Appendix  A:  Tables  from  the  Item-­level  Analyses  of  the  Catholic  School  –  Program  Effectiveness  Surveys  

Table  A.1.  Item  Response  Percentages  and  Intraclass  Correlation  Coefficients  (ICC)  for  the  Catholic  School  Program  Effectiveness  Survey  of  Adults  

Subscale    Item  

ICC  Strongly  Disagree  

     Strongly  Agree  

Don't  Know  

Skipped  

Mission  and  Catholic  Identity                  1. Everyone  in  the  school  community  –  

administrators,  faculty  and  staff,  students,  parents/guardians,  alums,  and  supporters  –  knows  and  understands  the  school’s  mission.  

0.08   2.2%   3.7%   12.7%   35.5%   45.8%   2.8%   8.2%  

2. Everything  we  do  in  our  school  is  guided  and  directed  by  our  mission.  

0.09   2.2%   4.7%   14.1%   34.7%   44.2%   2.4%   8.4%  

3. Our  school  mission  clearly  expresses  a  commitment  to  Catholic  identity.  

0.08   1.4%   2.1%   5.9%   25.2%   65.3%   1.4%   8.6%  

4. Our  school  provides  an  academically  rigorous  Catholic  religion  program,  taught  by  qualified  teachers.  

0.09   2.5%   3.8%   10.0%   29.3%   54.4%   1.4%   8.5%  

5. In  all  subjects,  teachers  help  students  think  critically  and  ethically  about  the  world  around  them,  using  the  lens  of  Gospel  values  and  Catholic  doctrine  and  beliefs.  

0.06   2.4%   4.4%   14.3%   36.0%   42.8%   3.0%   8.4%  

6. Our  school  provides  opportunities  outside  the  classroom  for  students  to  participate  in  service  activities  for  social  justice.  

0.09   1.7%   3.5%   10.3%   28.7%   55.9%   2.2%   8.5%  

7. Administrators,  faculty,  and  staff  serve  as  role  models  of  faith  and  service  to  students.  

0.09   3.2%   6.7%   16.6%   30.2%   43.3%   5.6%   8.4%  

8. Our  school  provides  opportunities  outside  the  classroom  for  student  faith  formation,  and  participation  in  retreats,  prayer,  mass,  sacraments,  and  other  spiritual  experiences.  

0.06   2.6%   4.1%   11.8%   31.4%   50.1%   1.0%   8.7%  

9. Our  school  provides  opportunities  for  faith  formation  for  faculty  and  staff.  

0.05   1.2%   3.6%   11.9%   31.0%   52.3%   19.7%   8.8%  

10. Our  school  provides  opportunities  for  faith  formation  for  parents/guardians  and  other  adult  members  of  the  school  community.  

0.06   4.0%   8.6%   20.6%   29.9%   36.9%   7.1%   8.8%  

11. Our  school  helps  parents/guardians  support  the  faith  life  of  their  child.  

0.06   2.0%   4.7%   13.8%   30.6%   48.8%   1.2%   8.8%  

Page 55: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  55      

12. Our  school  provides  opportunities  for  adult  members  of  the  school  community  to  participate  in  service  activities  for  social  justice.  

0.03   4.5%   8.2%   20.9%   30.3%   36.1%   10.8%   8.8%  

13. Every  adult  in  the  school  supports  the  faith  life  of  the  school  community.  

0.05   4.4%   8.0%   19.2%   32.1%   36.3%   9.4%   8.9%  

Governance  and  Leadership                  14. There  is  a  person  or  group  (such  as  a  

pastor  or  a  board)  who  collaborates  with  the  school  administration  to  make  or  recommend  decisions  for  the  success  of  the  school.  

0.08   2.1%   2.8%   9.6%   27.7%   57.9%   10.2%   11.6%  

15. A  person  or  group  (such  as  a  pastor  or  a  board),  in  collaboration  with  the  school  administration,  takes  responsibility  for  monitoring  that  the  school  is  faithful  to  its  mission,  academically  excellent  and  sound  in  its  business  decisions.  

0.09   2.8%   3.3%   12.1%   29.3%   52.5%   12.0%   11.8%  

16. Our  school  administration  effectively  carries  out  its  responsibilities  in  the  areas  of  faith  formation  and  instructional  leadership.  

0.09   2.5%   3.9%   10.7%   31.0%   51.9%   2.9%   11.8%  

17. Our  school  administration  has  authority  to  realize  and  implement  the  school’s  mission  and  vision.  

0.07   1.3%   2.8%   9.4%   28.6%   57.9%   5.6%   12.2%  

18. Our  school  administration  involves  all  members  of  the  school  community  to  ensure  a  school  culture  that  embodies  the  mission  and  vision.  

0.07   4.4%   5.9%   14.3%   28.1%   47.2%   5.0%   12.1%  

19. Our  school  administration  takes  responsibility  for  development  and  oversight  of  school  programs,  personnel,  and  school  operations.  

0.09   3.2%   4.1%   10.9%   28.1%   53.8%   3.6%   12.1%  

20. Our  school  has  a  strong  culture  of  collaboration  within  the  school  at  all  levels  to  advance  excellence.  

0.10   4.4%   5.8%   13.1%   29.6%   47.0%   4.3%   12.1%  

Academic  Excellence                  21. Our  school  has  a  clearly  articulated  

rigorous  curriculum  infused  with  Gospel  values,  preparing  students  for  life  and  work.  

0.07   2.4%   3.9%   13.3%   31.7%   48.7%   1.7%   16.6%  

22. In  classes  in  our  school,  students  spend  most  of  the  time  solving  problems,  discussing  ideas,  creating  their  own  work,  reading,  writing,  speaking,  and  researching.  

0.04   2.7%   5.0%   14.3%   33.7%   44.3%   4.3%   16.7%  

23. Curriculum  and  instruction  in  our  school  prepares  students  to  be  capable  and  critical  users  of  media  and  technology.  

0.11   3.6%   6.2%   13.7%   33.3%   43.1%   1.6%   16.7%  

24. Teachers  use  effective  instruction  to  deliver  the  curriculum.  

0.08   2.5%   5.1%   13.3%   34.5%   44.6%   2.0%   16.8%  

Page 56: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  56      

25. At  our  school,  teachers  use  different  teaching  approaches  to  meet  the  diverse  needs  of  all  students.  

0.05   5.5%   7.1%   14.7%   31.5%   41.1%   6.9%   16.8%  

26. At  our  school,  teachers  collaborate  systematically  and  regularly  in  order  to  increase  student  achievement  and  improve  teaching  effectiveness.  

0.06   3.7%   5.8%   14.9%   31.4%   44.2%   10.6%   17.0%  

27. At  our  school,  all  administrators,  faculty  and  staff  engage  in  ongoing  professional  development.  

0.09   2.6%   3.2%   11.4%   30.0%   52.8%   13.3%   16.7%  

28. Our  school  uses  standardized  and  teacher-­‐developed  assessments  to  document  student  learning  and  report  the  outcomes  to  parents/guardians.  

0.07   2.1%   3.4%   9.6%   29.7%   55.1%   5.2%   16.9%  

29. Our  school  uses  the  results  of  standardized  and  teacher-­‐developed  assessments  to  improve  the  curriculum  and  increase  learning.  

0.09   3.2%   4.2%   13.1%   29.1%   50.4%   12.3%   16.8%  

30. Teachers  vary  the  types  of  assessments  they  use  to  monitor  individual  and  class-­‐wide  student  learning.  

0.05   4.7%   6.0%   14.8%   33.1%   41.4%   13.0%   17.1%  

31. Our  school  communicates  how  well  students  are  achieving  in  comparison  to  similar  students  locally  and/or  nationally.  

0.09   8.5%   8.8%   16.5%   27.6%   38.6%   5.4%   17.0%  

32. Our  school  provides  programs  and  services  that  help  students  successfully  complete  the  school  program  (for  example,  guidance  and  resource  programs).  

0.09   5.2%   7.2%   16.2%   28.5%   42.9%   4.6%   16.9%  

33. Our  school  provides  enriching  programs  for  students  to  develop  their  gifts  and  talents,  and  enhance  their  creative,  artistic,  social/emotional,  physical,  and  spiritual  potential.  

0.08   5.5%   7.0%   15.7%   28.2%   43.5%   1.7%   17.1%  

34. Our  school  provides  opportunities  for  parents/guardians  to  be  involved  in  the  education  of  their  children.  

0.05   4.0%   5.8%   13.1%   27.6%   49.5%   1.3%   17.0%  

Operational  Vitality                  35. Our  school’s  financial  plan  is  the  result  

of  a  collaborative  process  including  expert  advisors.  

0.11   5.2%   6.2%   14.7%   33.3%   40.5%   25.1%   18.0%  

36. Our  school  consistently  shares  its  financial  plan  with  the  school  community.  

0.11   10.7%   11.2%   19.7%   26.6%   31.8%   10.7%   18.0%  

37. Our  school  leaders  take  responsibility  for  ensuring  that  the  financial  plans  and  budgets  are  implemented  using  best  practices.  

0.10   4.3%   5.9%   14.5%   31.5%   43.7%   19.0%   18.1%  

38. Our  school  treats  all  personnel  with  consistency,  fairness,  and  justice.  

0.08   6.3%   6.6%   13.3%   28.5%   45.2%   14.1%   18.0%  

Page 57: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  57      

39. Our  school  maintains  and  shares  plans  for  managing  facilities  and  equipment.  

0.10   7.9%   8.7%   17.3%   29.4%   36.7%   12.7%   18.2%  

40. Our  school  maintains  and  shares  a  technology  management  plan.  

0.08   8.2%   9.0%   18.1%   27.9%   36.8%   14.9%   18.4%  

41. Our  school’s  facilities,  equipment,  and  technology  management  plans  are  designed  to  enhance  teaching  and  learning.  

0.11   3.9%   5.8%   16.3%   31.0%   43.0%   11.0%   18.2%  

42. Our  school  has  an  institutional  advancement  plan,  based  on  our  mission,  which  uses  current  and  effective  strategies  for  communications,  marketing,  enrollment  management,  and  development.  

0.11   5.7%   6.0%   16.5%   31.0%   40.8%   20.8%   18.2%  

Note.  Percentages  for  Never  through  Always  options  are  mutually  exclusive  and  exhaustive  and  sum  to  100%;  participants  who  selected  Don't  Know  or  skipped  the  question  were  excluded  from  the  denominator  in  calculation  of  these  percentages.  The  percentages  reported  for  Don't  Know  or  skipped  responses  include  all  participants  in  the  denominator.  

Table  A.2.  School-­level  Statistics  for  the  Catholic  School  Program  Effectiveness  Survey  of  Adults  

Subscale    Item  

Average  Rating  

Standard  Deviation  

Minimum    Rating  

Maximum  Rating  

25th  Percentile  

Median  75th  

Percentile  

Mission  and  Catholic  Identity                1. Everyone  in  the  school  community  –  

administrators,  faculty  and  staff,  students,  parents/guardians,  alums,  and  supporters  –  knows  and  understands  the  school’s  mission.  

4.15   0.35   2.90   4.67   4.00   4.22   4.41  

2. Everything  we  do  in  our  school  is  guided  and  directed  by  our  mission.  

4.11   0.36   2.82   4.57   3.98   4.14   4.35  

3. Our  school  mission  clearly  expresses  a  commitment  to  Catholic  identity.  

4.51   0.25   3.75   4.89   4.40   4.52   4.69  

4. Our  school  provides  an  academically  rigorous  Catholic  religion  program,  taught  by  qualified  teachers.  

4.29   0.34   3.25   5.00   4.13   4.29   4.53  

5. In  all  subjects,  teachers  help  students  think  critically  and  ethically  about  the  world  around  them,  using  the  lens  of  Gospel  values  and  Catholic  doctrine  and  beliefs.  

4.13   0.33   3.00   4.75   4.00   4.15   4.37  

6. Our  school  provides  opportunities  outside  the  classroom  for  students  to  participate  in  service  activities  for  social  justice.  

4.37   0.29   3.75   4.92   4.15   4.37   4.57  

7. Administrators,  faculty,  and  staff  serve  as  role  models  of  faith  and  service  to  students.  

4.04   0.33   3.39   4.80   3.84   4.01   4.29  

8. Our  school  provides  opportunities  outside  the  classroom  for  student  faith  formation,  and  participation  in  retreats,  prayer,  mass,  sacraments,  and  other  spiritual  

4.25   0.31   3.33   4.76   4.10   4.25   4.45  

Page 58: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  58      

experiences.  

9. Our  school  provides  opportunities  for  faith  formation  for  faculty  and  staff.  

4.31   0.26   3.57   4.83   4.17   4.32   4.49  

10. Our  school  provides  opportunities  for  faith  formation  for  parents/guardians  and  other  adult  members  of  the  school  community.  

3.88   0.34   3.14   4.75   3.65   3.84   4.10  

11. Our  school  helps  parents/guardians  support  the  faith  life  of  their  child.  

4.20   0.30   3.43   4.82   4.02   4.22   4.42  

12. Our  school  provides  opportunities  for  adult  members  of  the  school  community  to  participate  in  service  activities  for  social  justice.  

3.85   0.30   3.29   4.67   3.59   3.86   4.05  

13. Every  adult  in  the  school  supports  the  faith  life  of  the  school  community.  

3.90   0.33   2.91   4.83   3.73   3.89   4.10  

Governance  and  Leadership                

14. There  is  a  person  or  group  (such  as  a  pastor  or  a  board)  who  collaborates  with  the  school  administration  to  make  or  recommend  decisions  for  the  success  of  the  school.  

4.34   0.31   2.77   4.76   4.27   4.39   4.48  

15. A  person  or  group  (such  as  a  pastor  or  a  board),  in  collaboration  with  the  school  administration,  takes  responsibility  for  monitoring  that  the  school  is  faithful  to  its  mission,  academically  excellent  and  sound  in  its  business  decisions.  

4.22   0.36   2.76   4.82   4.06   4.29   4.43  

16. Our  school  administration  effectively  carries  out  its  responsibilities  in  the  areas  of  faith  formation  and  instructional  leadership.  

4.25   0.35   2.83   4.75   4.09   4.29   4.50  

17. Our  school  administration  has  authority  to  realize  and  implement  the  school’s  mission  and  vision.  

4.37   0.28   3.25   4.81   4.28   4.41   4.54  

18. Our  school  administration  involves  all  members  of  the  school  community  to  ensure  a  school  culture  that  embodies  the  mission  and  vision.  

4.08   0.37   2.60   4.80   3.93   4.08   4.30  

19. Our  school  administration  takes  responsibility  for  development  and  oversight  of  school  programs,  personnel,  and  school  operations.  

4.23   0.38   2.92   4.82   4.11   4.24   4.50  

20. Our  school  has  a  strong  culture  of  collaboration  within  the  school  at  all  levels  to  advance  excellence.  

4.08   0.40   2.83   4.75   3.91   4.09   4.36  

Academic  Excellence                

21. Our  school  has  a  clearly  articulated  rigorous  curriculum  infused  with  Gospel  values,  preparing  students  for  life  and  work.  

4.19   0.31   3.18   4.82   4.00   4.21   4.43  

22. In  classes  in  our  school,  students  spend  most  of  the  time  solving  problems,  

4.12   0.29   3.09   4.68   3.97   4.14   4.32  

Page 59: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  59      

discussing  ideas,  creating  their  own  work,  reading,  writing,  speaking,  and  researching.  

23. Curriculum  and  instruction  in  our  school  prepares  students  to  be  capable  and  critical  users  of  media  and  technology.  

4.02   0.40   2.91   4.72   3.78   4.10   4.29  

24. Teachers  use  effective  instruction  to  deliver  the  curriculum.  

4.14   0.37   2.50   4.76   3.99   4.15   4.34  

25. At  our  school,  teachers  use  different  teaching  approaches  to  meet  the  diverse  needs  of  all  students.  

3.99   0.38   2.70   4.67   3.83   4.02   4.20  

26. At  our  school,  teachers  collaborate  systematically  and  regularly  in  order  to  increase  student  achievement  and  improve  teaching  effectiveness.  

4.07   0.39   2.64   4.80   3.94   4.13   4.28  

27. At  our  school,  all  administrators,  faculty  and  staff  engage  in  ongoing  professional  development.  

4.26   0.36   2.92   4.80   4.15   4.36   4.47  

28. Our  school  uses  standardized  and  teacher-­‐developed  assessments  to  document  student  learning  and  report  the  outcomes  to  parents/guardians.  

4.33   0.32   3.40   4.91   4.20   4.39   4.53  

29. Our  school  uses  the  results  of  standardized  and  teacher-­‐developed  assessments  to  improve  the  curriculum  and  increase  learning.  

4.20   0.39   3.00   5.00   4.05   4.25   4.41  

30. Teachers  vary  the  types  of  assessments  they  use  to  monitor  individual  and  class-­‐wide  student  learning.  

4.03   0.36   2.91   5.00   3.88   4.08   4.21  

31. Our  school  communicates  how  well  students  are  achieving  in  comparison  to  similar  students  locally  and/or  nationally.  

3.82   0.46   2.42   4.50   3.60   3.91   4.18  

32. Our  school  provides  programs  and  services  that  help  students  successfully  complete  the  school  program  (for  example,  guidance  and  resource  programs).  

3.93   0.41   2.55   4.73   3.75   3.94   4.21  

33. Our  school  provides  enriching  programs  for  students  to  develop  their  gifts  and  talents,  and  enhance  their  creative,  artistic,  social/emotional,  physical,  and  spiritual  potential.  

3.93   0.36   2.91   4.64   3.73   3.98   4.20  

34. Our  school  provides  opportunities  for  parents/guardians  to  be  involved  in  the  education  of  their  children.  

4.13   0.38   2.64   4.75   4.00   4.21   4.35  

Operational  Vitality                

35. Our  school’s  financial  plan  is  the  result  of  a  collaborative  process  including  expert  advisors.  

3.92   0.43   2.48   4.67   3.67   3.95   4.26  

36. Our  school  consistently  shares  its  financial  plan  with  the  school  community.  

3.54   0.48   2.50   4.65   3.20   3.59   3.87  

37. Our  school  leaders  take  responsibility  for  ensuring  that  the  financial  plans  and  budgets  are  implemented  using  best  

4.02   0.41   2.52   4.79   3.80   4.10   4.27  

Page 60: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  60      

practices.  

38. Our  school  treats  all  personnel  with  consistency,  fairness,  and  justice.  

3.96   0.44   2.55   4.80   3.82   4.05   4.26  

39. Our  school  maintains  and  shares  plans  for  managing  facilities  and  equipment.  

3.76   0.44   2.31   4.76   3.61   3.81   4.00  

40. Our  school  maintains  and  shares  a  technology  management  plan.  

3.75   0.44   2.57   4.50   3.52   3.79   4.13  

41. Our  school’s  facilities,  equipment,  and  technology  management  plans  are  designed  to  enhance  teaching  and  learning.  

4.00   0.40   3.08   4.67   3.74   4.11   4.31  

42. Our  school  has  an  institutional  advancement  plan,  based  on  our  mission,  which  uses  current  and  effective  strategies  for  communications,  marketing,  enrollment  management,  and  development.  

3.89   0.45   2.45   4.67   3.67   3.95   4.21  

 

Table  A.3.  Standardized  Factor  Loadings  for  the  1-­factor  CFA  Model  (School-­level)  of  the  Catholic  School  Program  Effectiveness  Survey  of  Adults  

Full-­‐Scale  1-­‐Factor  Model   Subscale  1-­‐Factor  Models  

Subscale    Item)  

Standardized  Factor  Loading  

Standard  Error  

Standardized  Factor  Loading  

Standard  Error  

Mission  and  Catholic  Identity          1. Everyone  in  the  school  community  –  

administrators,  faculty  and  staff,  students,  parents/guardians,  alums,  and  supporters  –  knows  and  understands  the  school’s  mission.  

0.90   0.04   0.89   0.05  

2. Everything  we  do  in  our  school  is  guided  and  directed  by  our  mission.  

0.96   0.02   0.97   0.03  

3. Our  school  mission  clearly  expresses  a  commitment  to  Catholic  identity.  

0.94   0.03   0.99   0.03  

4. Our  school  provides  an  academically  rigorous  Catholic  religion  program,  taught  by  qualified  teachers.  

0.93   0.04   0.92   0.04  

5. In  all  subjects,  teachers  help  students  think  critically  and  ethically  about  the  world  around  them,  using  the  lens  of  Gospel  values  and  Catholic  doctrine  and  beliefs.  

0.87   0.05   0.91   0.04  

6. Our  school  provides  opportunities  outside  the  classroom  for  students  to  participate  in  service  activities  for  social  justice.  

0.68   0.08   0.75   0.07  

7. Administrators,  faculty,  and  staff  serve  as  role  models  of  faith  and  service  to  students.  

0.60   0.10   0.58   0.11  

Page 61: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  61      

8. Our  school  provides  opportunities  outside  the  classroom  for  student  faith  formation,  and  participation  in  retreats,  prayer,  mass,  sacraments,  and  other  spiritual  experiences.  

0.93   0.03   0.95   0.03  

9. Our  school  provides  opportunities  for  faith  formation  for  faculty  and  staff.  

0.82   0.07   0.86   0.06  

10. Our  school  provides  opportunities  for  faith  formation  for  parents/guardians  and  other  adult  members  of  the  school  community.  

0.58   0.09   0.67   0.08  

11. Our  school  helps  parents/guardians  support  the  faith  life  of  their  child.  

0.82   0.06   0.95   0.02  

12. Our  school  provides  opportunities  for  adult  members  of  the  school  community  to  participate  in  service  activities  for  social  justice.  

0.85   0.07   0.88   0.07  

13. Every  adult  in  the  school  supports  the  faith  life  of  the  school  community.  

0.91   0.04   0.97   0.03  

Governance  and  Leadership          14. There  is  a  person  or  group  (such  as  a  

pastor  or  a  board)  who  collaborates  with  the  school  administration  to  make  or  recommend  decisions  for  the  success  of  the  school.  

0.80   0.07   0.86   0.05  

15. A  person  or  group  (such  as  a  pastor  or  a  board),  in  collaboration  with  the  school  administration,  takes  responsibility  for  monitoring  that  the  school  is  faithful  to  its  mission,  academically  excellent  and  sound  in  its  business  decisions.  

0.89   0.04   0.94   0.03  

16. Our  school  administration  effectively  carries  out  its  responsibilities  in  the  areas  of  faith  formation  and  instructional  leadership.  

0.95   0.02   0.96   0.02  

17. Our  school  administration  has  authority  to  realize  and  implement  the  school’s  mission  and  vision.  

0.98   0.03   0.96   0.02  

18. Our  school  administration  involves  all  members  of  the  school  community  to  ensure  a  school  culture  that  embodies  the  mission  and  vision.  

0.97   0.02   0.98   0.02  

19. Our  school  administration  takes  responsibility  for  development  and  oversight  of  school  programs,  personnel,  and  school  operations.  

0.98   0.02   0.98   0.02  

20. Our  school  has  a  strong  culture  of  collaboration  within  the  school  at  all  levels  to  advance  excellence.  

0.98   0.02   0.96   0.02  

Academic  Excellence          

21. Our  school  has  a  clearly  articulated  rigorous  curriculum  infused  with  Gospel  values,  preparing  students  for  life  and  

0.89   0.04   0.89   0.03  

Page 62: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  62      

work.  22. In  classes  in  our  school,  students  spend  

most  of  the  time  solving  problems,  discussing  ideas,  creating  their  own  work,  reading,  writing,  speaking,  and  researching.  

0.85   0.06   0.93   0.04  

23. Curriculum  and  instruction  in  our  school  prepares  students  to  be  capable  and  critical  users  of  media  and  technology.  

0.86   0.05   0.86   0.05  

24. Teachers  use  effective  instruction  to  deliver  the  curriculum.  

0.89   0.04   0.94   0.03  

25. At  our  school,  teachers  use  different  teaching  approaches  to  meet  the  diverse  needs  of  all  students.  

0.88   0.05   0.96   0.03  

26. At  our  school,  teachers  collaborate  systematically  and  regularly  in  order  to  increase  student  achievement  and  improve  teaching  effectiveness.  

0.91   0.04   0.94   0.03  

27. At  our  school,  all  administrators,  faculty  and  staff  engage  in  ongoing  professional  development.  

0.90   0.04   0.85   0.05  

28. Our  school  uses  standardized  and  teacher-­‐developed  assessments  to  document  student  learning  and  report  the  outcomes  to  parents/guardians.  

0.90   0.04   0.94   0.03  

29. Our  school  uses  the  results  of  standardized  and  teacher-­‐developed  assessments  to  improve  the  curriculum  and  increase  learning.  

0.87   0.05   0.95   0.03  

30. Teachers  vary  the  types  of  assessments  they  use  to  monitor  individual  and  class-­‐wide  student  learning.  

0.86   0.05   0.96   0.03  

31. Our  school  communicates  how  well  students  are  achieving  in  comparison  to  similar  students  locally  and/or  nationally.  

0.77   0.06   0.87   0.05  

32. Our  school  provides  programs  and  services  that  help  students  successfully  complete  the  school  program  (for  example,  guidance  and  resource  programs).  

0.71   0.07   0.71   0.08  

33. Our  school  provides  enriching  programs  for  students  to  develop  their  gifts  and  talents,  and  enhance  their  creative,  artistic,  social/emotional,  physical,  and  spiritual  potential.  

0.75   0.07   0.73   0.07  

34. Our  school  provides  opportunities  for  parents/guardians  to  be  involved  in  the  education  of  their  children.  

0.88   0.04   0.86   0.05  

Operational  Vitality          35. Our  school’s  financial  plan  is  the  result  of  

a  collaborative  process  including  expert  advisors.  

0.93   0.04   0.94   0.03  

36. Our  school  consistently  shares  its   0.78   0.06   0.88   0.04  

Page 63: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  63      

financial  plan  with  the  school  community.  

37. Our  school  leaders  take  responsibility  for  ensuring  that  the  financial  plans  and  budgets  are  implemented  using  best  practices.  

0.96   0.02   0.96   0.02  

38. Our  school  treats  all  personnel  with  consistency,  fairness,  and  justice.  

0.83   0.05   0.78   0.06  

39. Our  school  maintains  and  shares  plans  for  managing  facilities  and  equipment.  

0.89   0.03   0.97   0.01  

40. Our  school  maintains  and  shares  a  technology  management  plan.  

0.84   0.06   0.95   0.03  

41. Our  school’s  facilities,  equipment,  and  technology  management  plans  are  designed  to  enhance  teaching  and  learning.  

0.83   0.06   0.93   0.04  

42. Our  school  has  an  institutional  advancement  plan,  based  on  our  mission,  which  uses  current  and  effective  strategies  for  communications,  marketing,  enrollment  management,  and  development.  

0.89   0.04   0.95   0.02  

aThe  error  variance  for  this  item  was  original  estimated  in  the  subscale  CFA  analysis  as  a  negative  number  and  therefore  was  subsequently  constrained  to  a  small  positive  constant  (i.e.,  .001).    

Table  A.4.  Item  Response  Frequencies  and  Intraclass  Correlation  Coefficients  (ICC)  for  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5-­8  Grades)  

Subscale    Item  

ICC  Strongly  Disagree  

     Strongly  Agree  

Don't  Know  

Skipped  

Mission  and  Catholic  Identity                  1. Everyone  in  the  school  community  –  

the  principal,  teachers  and  staff,  students,  and  parents/guardians  –  knows  and  understands  the  school’s  mission.  

0.04   2.8%   4.9%   16.7%   36.3%   39.4%   9.6%   3.2%  

2. Our  school’s  mission  affects  everything  we  do.  

0.03   5.4%   8.8%   20.9%   35.2%   29.7%   11.6%   3.6%  

3. Our  school’s  mission  statement  clearly  says  that  we  are  a  Catholic  school.  

0.09   1.1%   2.1%   5.8%   17.8%   73.2%   4.9%   4.7%  

4. Religion  class  in  our  school  is  very  important,  and  is  taught  by  knowledgeable  teachers.  

0.11   2.9%   3.7%   7.3%   20.3%   65.7%   0.8%   3.9%  

5. In  all  subjects,  teachers  help  students  think  about  how  Gospel  values  and  Catholic  beliefs  can  help  to  make  the  world  a  better  place.  

0.06   8.1%   13.3%   23.8%   29.0%   25.8%   3.7%   3.8%  

6. Our  school  provides  opportunities  outside  the  classroom  for  students  to  develop  their  faith  and  participate  in  Mass,  sacraments  and  prayer.  

0.06   3.1%   5.3%   11.2%   27.4%   53.0%   2.9%   4.4%  

Page 64: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  64      

7. Our  school  provides  opportunities  outside  the  classroom  for  students  to  do  service  on  behalf  of  the  poor  and  those  in  need.  

0.04   5.2%   7.5%   15.3%   28.8%   43.3%   4.9%   3.9%  

8. In  our  school,  adults  do  service  on  behalf  of  the  poor  and  those  in  need.  

0.02   5.1%   8.4%   18.3%   33.3%   35.0%   10.5%   4.6%  

9. The  principal,  teachers,  and  staff  serve  as  role  models  of  faith  and  service  to  students.  

0.06   6.3%   7.1%   14.9%   27.4%   44.3%   1.7%   3.5%  

Governance  and  Leadership                  10. The  principal  makes  sure  that  we  have  

a  successful  school.  0.16   3.8%   3.8%   7.6%   22.9%   61.8%   1.8%   3.6%  

11. The  principal  involves  everyone  in  the  school  community  to  make  sure  that  everything  we  do  furthers  the  school’s  mission  and  vision.  

0.08   5.2%   6.2%   15.4%   33.2%   40.1%   6.7%   3.8%  

12. The  principal  finds  ways  to  help  teachers,  staff,  students,  and  parents  work  together  in  striving  for  excellence  in  all  aspects  of  the  school.  

0.09   5.7%   6.3%   13.1%   28.7%   46.2%   5.4%   3.9%  

Academic  Excellence                  13. Our  school  prepares  students  for  the  

best  high  schools.  0.08   3.7%   4.7%   11.5%   29.0%   51.1%   2.9%   4.2%  

14. Our  school  prepares  students  to  use  and  judge  media  and  technology.  

0.07   6.8%   8.3%   19.7%   33.7%   31.4%   7.0%   4.7%  

15. In  our  classes,  students  spend  most  of  the  time  solving  problems,  discussing  ideas,  creating  their  own  work,  reading,  writing,  speaking,  and  researching.  

0.02   5.6%   7.5%   15.2%   29.1%   42.5%   1.3%   4.7%  

16. Our  teachers  do  a  good  job  of  helping  all  students  learn.  

0.09   5.0%   7.1%   14.8%   28.1%   45.0%   0.8%   4.8%  

17. At  our  school,  teachers  use  many  different  approaches  in  their  classrooms  to  meet  the  different  needs  of  all  students.  

0.04   6.1%   8.3%   15.3%   28.6%   41.7%   2.6%   4.8%  

18. At  our  school,  teachers  work  together  to  help  each  other  become  better  teachers.  

0.06   6.0%   7.3%   16.7%   29.3%   40.6%   13.8%   4.7%  

19. At  our  school,  student  work  is  graded  fairly.  

0.07   8.5%   7.8%   14.3%   26.8%   42.6%   3.1%   4.9%  

20. At  our  school,  teachers  use  student  work  and  student  test  results  to  improve  how  they  teach.  

0.04   8.4%   9.2%   15.8%   30.2%   36.4%   16.4%   5.5%  

21. At  our  school,  teachers  find  many  different  ways  to  see  how  much  students  have  learned.  

0.04   5.7%   7.7%   16.5%   30.1%   40.1%   5.9%   5.2%  

22. Our  school  offers  guidance  and  resources  (such  as  counselors,  tutors  and  special  teachers)  to  help  students  be  successful.  

0.17   6.9%   6.8%   12.1%   22.9%   51.2%   4.0%   4.7%  

Page 65: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  65      

23. Our  school  offers  programs  and  activities  (such  as  sports,  drama,  clubs,  or  band)  for  students  to  develop  their  gifts  and  talents.  

0.21   3.6%   4.1%   7.7%   16.9%   67.7%   1.0%   5.0%  

24. Our  school  invites  parents/guardians  to  be  involved  in  the  school.  

0.07   3.2%   3.8%   11.2%   25.0%   56.8%   2.4%   5.2%  

Operational  Vitality                  25. Our  school  provides  the  space,  

equipment  and  technology  we  need  to  learn.  

0.07   5.6%   7.0%   14.5%   31.6%   41.3%   1.5%   4.8%  

26. Our  school  uses  different  ways  to  keep  parents/guardians  informed  about  all  that  is  happening  in  our  school.  

0.04   4.5%   5.0%   11.9%   30.7%   47.9%   2.6%   5.0%  

27. Our  school  does  a  good  job  of  attracting  new  students  to  our  school.  

0.08   9.0%   7.7%   17.6%   27.9%   37.8%   6.9%   5.0%  

 

Table  A.5.  School-­level  Statistics  for  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5-­8  Grades)  

Subscale    Item  

Average  Rating  

Standard  Deviation  

Minimum    Rating  

Maximum  Rating  

25th  Percentile  

Median  75th  

Percentile  

Mission  and  Catholic  Identity                1. Everyone  in  the  school  community  –  the  

principal,  teachers  and  staff,  students,  and  parents/guardians  –  knows  and  understands  the  school’s  mission.  

4.03   0.27   3.33   4.51   3.87   4.10   4.21  

2. Our  school’s  mission  affects  everything  we  do.  

3.83   0.32   3.22   4.75   3.60   3.83   4.00  

3. Our  school’s  mission  statement  clearly  says  that  we  are  a  Catholic  school.  

4.61   0.22   4.07   5.00   4.51   4.63   4.78  

4. Religion  class  in  our  school  is  very  important,  and  is  taught  by  knowledgeable  teachers.  

4.43   0.31   3.50   5.00   4.23   4.43   4.66  

5. In  all  subjects,  teachers  help  students  think  about  how  Gospel  values  and  Catholic  beliefs  can  help  to  make  the  world  a  better  place.  

3.60   0.43   2.50   4.75   3.31   3.53   3.82  

6. Our  school  provides  opportunities  outside  the  classroom  for  students  to  develop  their  faith  and  participate  in  Mass,  sacraments  and  prayer.  

4.23   0.28   3.50   4.70   4.01   4.28   4.43  

7. Our  school  provides  opportunities  outside  the  classroom  for  students  to  do  service  on  behalf  of  the  poor  and  those  in  need.  

3.98   0.28   3.32   4.54   3.80   4.00   4.15  

8. In  our  school,  adults  do  service  on  behalf  of  the  poor  and  those  in  need.  

3.90   0.25   3.41   4.75   3.71   3.88   4.07  

9. The  principal,  teachers,  and  staff  serve  as  role  models  of  faith  and  service  to  students.  

3.95   0.38   3.00   4.64   3.77   3.98   4.23  

Page 66: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  66      

Governance  and  Leadership                

10. The  principal  makes  sure  that  we  have  a  successful  school.  

4.34   0.48   2.71   5.00   4.12   4.47   4.69  

11. The  principal  involves  everyone  in  the  school  community  to  make  sure  that  everything  we  do  furthers  the  school’s  mission  and  vision.  

3.97   0.44   2.71   4.80   3.74   3.99   4.32  

12. The  principal  finds  ways  to  help  teachers,  staff,  students,  and  parents  work  together  in  striving  for  excellence  in  all  aspects  of  the  school.  

4.04   0.47   2.71   4.80   3.81   4.15   4.32  

Academic  Excellence                

13. Our  school  prepares  students  for  the  best  high  schools.  

4.17   0.35   2.93   4.70   3.96   4.24   4.43  

14. Our  school  prepares  students  to  use  and  judge  media  and  technology.  

3.72   0.38   2.75   4.45   3.43   3.75   4.01  

15. In  our  classes,  students  spend  most  of  the  time  solving  problems,  discussing  ideas,  creating  their  own  work,  reading,  writing,  speaking,  and  researching.  

3.95   0.28   3.33   4.70   3.73   3.95   4.15  

16. Our  teachers  do  a  good  job  of  helping  all  students  learn.  

4.03   0.39   3.00   4.81   3.80   4.08   4.32  

17. At  our  school,  teachers  use  many  different  approaches  in  their  classrooms  to  meet  the  different  needs  of  all  students.  

3.93   0.37   2.86   4.60   3.69   3.98   4.11  

18. At  our  school,  teachers  work  together  to  help  each  other  become  better  teachers.  

3.88   0.44   2.27   4.70   3.68   3.97   4.17  

19. At  our  school,  student  work  is  graded  fairly.  

3.91   0.41   2.71   4.71   3.64   4.00   4.22  

20. At  our  school,  teachers  use  student  work  and  student  test  results  to  improve  how  they  teach.  

3.76   0.44   2.36   4.88   3.56   3.80   4.04  

21. At  our  school,  teachers  find  many  different  ways  to  see  how  much  students  have  learned.  

3.91   0.31   2.75   4.50   3.75   3.96   4.11  

22. Our  school  offers  guidance  and  resources  (such  as  counselors,  tutors  and  special  teachers)  to  help  students  be  successful.  

3.91   0.58   1.86   4.88   3.49   4.07   4.31  

23. Our  school  offers  programs  and  activities  (such  as  sports,  drama,  clubs,  or  band)  for  students  to  develop  their  gifts  and  talents.  

4.21   0.58   2.43   4.88   3.95   4.31   4.65  

24. Our  school  invites  parents/guardians  to  be  involved  in  the  school.  

4.26   0.35   2.71   4.78   4.10   4.35   4.47  

Operational  Vitality                

25. Our  school  provides  the  space,  equipment  and  technology  we  need  to  learn.  

3.89   0.37   2.75   4.62   3.70   3.92   4.09  

26. Our  school  uses  different  ways  to  keep  parents/guardians  informed  about  all  that  is  happening  in  our  school.  

4.13   0.31   2.86   4.60   3.98   4.18   4.33  

Page 67: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  67      

27. Our  school  does  a  good  job  of  attracting  new  students  to  our  school.  

3.70   0.45   2.40   4.50   3.54   3.79   3.96  

 

Table  A.6.  Standardized  Factor  Loadings  for  the  1-­factor  CFA  Model  (School-­level)  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (5-­8  Grades)  

Full-­‐Scale  1-­‐Factor  Model   Subscale  1-­‐Factor  Models  

Subscale    Item  

Standardized  Factor  Loading  

Standard  Error  

Standardized  Factor  Loading  

Standard  Error  

Mission  and  Catholic  Identity          1. Everyone  in  the  school  community  –  the  

principal,  teachers  and  staff,  students,  and  parents/guardians  –  knows  and  understands  the  school’s  mission.  

0.96   0.04   0.90   0.06  

2. Our  school’s  mission  affects  everything  we  do.  

0.77   0.09   0.73   0.11  

3. Our  school’s  mission  statement  clearly  says  that  we  are  a  Catholic  school.  

0.75   0.11   0.73   0.11  

4. Religion  class  in  our  school  is  very  important,  and  is  taught  by  knowledgeable  teachers.  

0.61   0.13   0.69   0.11  

5. In  all  subjects,  teachers  help  students  think  about  how  Gospel  values  and  Catholic  beliefs  can  help  to  make  the  world  a  better  place.  

0.71   0.09   0.78   0.10  

6. Our  school  provides  opportunities  outside  the  classroom  for  students  to  develop  their  faith  and  participate  in  Mass,  sacraments  and  prayer.  

0.75   0.09   0.89   0.06  

7. Our  school  provides  opportunities  outside  the  classroom  for  students  to  do  service  on  behalf  of  the  poor  and  those  in  need.  

0.55   0.10   0.58   0.10  

8. In  our  school,  adults  do  service  on  behalf  of  the  poor  and  those  in  need.  

0.70   0.12   0.81   0.08  

9. The  principal,  teachers,  and  staff  serve  as  role  models  of  faith  and  service  to  students.  

0.97   0.03   0.94   0.05  

Governance  and  Leadership          

10. The  principal  makes  sure  that  we  have  a  successful  school.  

0.84   0.07   1.00   0.04  

11. The  principal  involves  everyone  in  the  school  community  to  make  sure  that  everything  we  do  furthers  the  school’s  mission  and  vision.  

0.88   0.05   0.95   0.04  

12. The  principal  finds  ways  to  help  teachers,  staff,  students,  and  parents  work  together  in  striving  for  excellence  in  all  aspects  of  the  school.  

0.82   0.07   0.99   0.03  

Academic  Excellence          

13. Our  school  prepares  students  for  the  best   0.63   0.09   0.64   0.12  

Page 68: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  68      

high  schools.  

14. Our  school  prepares  students  to  use  and  judge  media  and  technology.  

0.40   0.15   0.45   0.15  

15. In  our  classes,  students  spend  most  of  the  time  solving  problems,  discussing  ideas,  creating  their  own  work,  reading,  writing,  speaking,  and  researching.  

0.88   0.08   0.99   0.07  

16. Our  teachers  do  a  good  job  of  helping  all  students  learn.  

0.88   0.05   0.93   0.04  

17. At  our  school,  teachers  use  many  different  approaches  in  their  classrooms  to  meet  the  different  needs  of  all  students.  

0.88   0.06   0.97   0.03  

18. At  our  school,  teachers  work  together  to  help  each  other  become  better  teachers.  

0.89   0.06   0.96   0.04  

19. At  our  school,  student  work  is  graded  fairly.  

0.85   0.07   0.78   0.09  

20. At  our  school,  teachers  use  student  work  and  student  test  results  to  improve  how  they  teach.  

0.84   0.06   0.88   0.06  

21. At  our  school,  teachers  find  many  different  ways  to  see  how  much  students  have  learned.  

0.94   0.04   0.97   0.03  

22. Our  school  offers  guidance  and  resources  (such  as  counselors,  tutors  and  special  teachers)  to  help  students  be  successful.  

0.34   0.16   0.28   0.19  

23. Our  school  offers  programs  and  activities  (such  as  sports,  drama,  clubs,  or  band)  for  students  to  develop  their  gifts  and  talents.  

0.34   0.16   0.26   0.19  

24. Our  school  invites  parents/guardians  to  be  involved  in  the  school.  

0.80   0.07   0.70   0.09  

Operational  Vitality          

25. Our  school  provides  the  space,  equipment  and  technology  we  need  to  learn.  

0.49   0.13   0.50   0.11  

26. Our  school  uses  different  ways  to  keep  parents/guardians  informed  about  all  that  is  happening  in  our  school.  

0.81   0.08   0.50   0.12  

27. Our  school  does  a  good  job  of  attracting  new  students  to  our  school.a  

0.75   0.07   1.00   0.00  

aThe  error  variance  for  this  item  was  original  estimated  in  the  subscale  CFA  analysis  as  a  negative  number  and  therefore  was  subsequently  constrained  to  a  small  positive  constant  (i.e.,  .001).    

Table  A.7.  Item  Response  Frequencies  for  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9-­12  Grades)  

Subscale    Item  

Strongly  Disagree  

     Strongly  Agree  

Don't  Know  

Skipped  

Mission  and  Catholic  Identity                

Page 69: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  69      

1. Everyone  in  the  school  community  –  administrators,  faculty  and  staff,  students,  parents/guardians,  alums,  and  supporters  –  knows  and  understands  the  school’s  mission.  

1.1%   3.2%   20.0%   38.9%   36.8%   6.1%   10.5%  

2. Everything  we  do  in  our  school  is  guided  and  directed  by  our  mission.  

2.0%   6.1%   17.3%   46.9%   27.6%   3.5%   10.5%  

3. Our  school  mission  clearly  expresses  a  commitment  to  Catholic  identity.  

0.0%   2.0%   10.0%   37.0%   51.0%   1.8%   10.5%  

4. Our  school  provides  an  academically  rigorous  Catholic  religion  program,  taught  by  qualified  teachers.  

1.0%   5.9%   10.9%   38.6%   43.6%   0.0%   11.4%  

5. In  all  subjects,  teachers  help  students  think  critically  and  ethically  about  the  world  around  them,  using  the  lens  of  Gospel  values  and  Catholic  doctrine  and  beliefs.  

3.0%   10.9%   23.8%   33.7%   28.7%   0.0%   11.4%  

6. Our  school  provides  opportunities  outside  the  classroom  for  student  faith  formation,  and  participation  in  retreats,  prayer,  mass,  sacraments,  and  other  spiritual  experiences.    

1.0%   2.0%   9.2%   30.6%   57.1%   3.5%   10.5%  

7. Our  school  provides  opportunities  outside  the  classroom  for  students  to  participate  in  service  activities  for  social  justice.    

0.0%   2.1%   11.3%   30.9%   55.7%   0.0%   0.0%  

8. Administrators,  faculty,  and  staff  serve  as  role  models  of  faith  and  service  to  students.  

4.0%   4.0%   13.1%   41.4%   37.4%   0.9%   12.3%  

9. Our  school  provides  opportunities  for  faith  formation  for  faculty  and  staff.  

1.1%   1.1%   17.0%   43.2%   37.5%   10.5%   12.3%  

10. Our  school  provides  opportunities  for  faith  formation  for  parents/guardians  and  other  adult  members  of  the  school  community.  

2.2%   3.3%   24.2%   37.4%   33.0%   8.8%   11.4%  

11. Our  school  provides  opportunities  for  adult  members  of  the  school  community  to  participate  in  service  activities  for  social  justice.  

1.1%   7.8%   24.4%   34.4%   32.2%   9.6%   11.4%  

12. Our  school  helps  parents/guardians  support  the  faith  life  of  their  child.  

6.5%   3.3%   18.5%   32.6%   39.1%   6.1%   13.2%  

13. Every  adult  in  the  school  supports  the  faith  life  of  the  school  community.  

6.5%   4.3%   20.4%   35.5%   33.3%   7.0%   11.4%  

Governance  and  Leadership                14. There  is  a  person  or  group  (such  as  a  

board  or  a  pastor)  who  collaborates  with  the  school  administration  to  make  or  recommend  decisions  for  the  success  of  the  school.  

2.2%   3.2%   9.7%   38.7%   46.2%   5.3%   13.2%  

15. Our  school  administration  takes  responsibility  for  development  and  oversight  of  school  programs,  personnel,  and  school  operations.  

2.1%   4.3%   8.5%   43.6%   41.5%   4.4%   13.2%  

Page 70: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  70      

16. Our  school  administration  involves  all  members  of  the  school  community  to  make  sure  that  everything  we  do  embodies  the  school’s  mission  and  vision.  

4.1%   3.1%   16.5%   37.1%   39.2%   1.8%   13.2%  

17. Our  school  administration  finds  ways  to  help  faculty,  staff,  students,  and  parents  collaborate  in  striving  for  excellence  in  all  aspects  of  the  school.  

2.1%   1.1%   13.7%   44.2%   38.9%   3.5%   13.2%  

Academic  Excellence                18. Our  school  has  a  challenging  

curriculum  infused  with  Gospel  values,  preparing  students  for  college,  work,  and  life.  

3.4%   7.9%   15.7%   41.6%   31.5%   0.0%   21.9%  

19. Curriculum  and  instruction  in  our  school  prepares  students  to  be  capable  and  critical  users  of  media  and  technology.  

4.5%   9.0%   20.2%   41.6%   24.7%   0.0%   21.6%  

20. In  our  classes,  students  spend  most  of  the  time  solving  problems,  discussing  ideas,  creating  their  own  work,  reading,  writing,  speaking,  and  researching.  

2.3%   11.4%   20.5%   33.0%   33.0%   0.0%   22.8%  

21. Teachers  use  effective  instruction  to  deliver  the  curriculum.  

5.7%   6.8%   22.7%   40.9%   23.9%   0.9%   21.9%  

22. At  our  school,  teachers  use  different  teaching  approaches  to  meet  the  diverse  needs  of  all  students.  

8.0%   4.6%   20.7%   37.9%   28.7%   0.9%   22.8%  

23. At  our  school,  teachers  work  together  to  help  each  other  become  better  teachers.  

5.2%   5.2%   13.0%   49.4%   27.3%   9.6%   22.8%  

24. At  our  school,  all  administrators,  faculty  and  staff  engage  in  ongoing  professional  development.  

1.3%   7.7%   14.1%   46.2%   30.8%   8.8%   22.8%  

25. At  our  school,  teachers  follow  school-­‐wide  policies  and  procedures  to  fairly  evaluate  and  communicate  student  performance.  

3.7%   4.9%   14.8%   42.0%   34.6%   7.0%   21.9%  

26. At  our  school,  teachers  work  together  to  use  data  on  student  performance  to  review  and  improve  courses  and  instruction.  

3.9%   5.2%   15.6%   39.0%   36.4%   8.8%   23.7%  

27. Teachers  vary  the  types  of  assessments  they  use  to  monitor  individual  and  class-­‐wide  student  learning.  

3.6%   2.4%   19.3%   45.8%   28.9%   5.3%   21.9%  

28. Our  school  communicates  how  well  students  are  achieving  in  comparison  to  similar  students  locally  and/or  nationally.  

7.4%   8.6%   18.5%   34.6%   30.9%   7.0%   21.9%  

Page 71: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  71      

29. Our  school  provides  programs  and  services  that  help  students  successfully  complete  the  school  program  (for  example,  guidance  and  resource  programs).  

2.4%   5.9%   11.8%   40.0%   40.0%   3.5%   21.9%  

30. Our  school  provides  co-­‐curricular  and  extra-­‐curricular  programs  for  students  to  develop  gifts  and  talents  and  enhance  their  creative,  artistic,  social/emotional,  physical,  and  spiritual  potential.  

1.2%   3.6%   16.7%   34.5%   44.0%   2.6%   23.7%  

31. Our  school  provides  opportunities  for  parents/guardians  to  be  involved  in  the  education  of  their  students.  

4.7%   4.7%   21.2%   40.0%   29.4%   3.5%   21.9%  

Operational  Vitality                32. Our  school  treats  everyone  who  works  

at  the  school  with  consistency,  fairness,  and  justice.  

6.9%   6.9%   17.2%   32.2%   36.8%   0.9%   22.8%  

33. Our  school  manages  facilities,  equipment,  and  technology  in  ways  that  enhance  teaching  and  learning.  

8.0%   4.6%   25.3%   36.8%   25.3%   0.9%   22.8%  

34. Our  school  uses  different  ways  to  communicate  all  that  is  happening  in  our  school  to  parents/guardians,  the  school  community  and  beyond.  

2.4%   4.8%   21.7%   42.2%   28.9%   2.6%   24.6%  

35. Our  school  does  a  good  job  of  attracting  new  students  to  our  school.  

7.1%   4.8%   17.9%   39.3%   31.0%   3.5%   22.8%  

 

Table  A.8.  Standardized  Factor  Loadings  for  the  1-­factor  CFA  Model  (Student-­level)  of  the  Catholic  School  Program  Effectiveness  Survey  of  Students  (9-­12  Grades)  

Full-­‐Scale  1-­‐Factor  Model   Subscale  1-­‐Factor  Models  

Subscale    Item    

Standardized  Factor  Loading  

Standard  Error  

Standardized  Factor  Loading  

Standard  Error  

Mission  and  Catholic  Identity          1. Everyone  in  the  school  community  –  

administrators,  faculty  and  staff,  students,  parents/guardians,  alums,  and  supporters  –  knows  and  understands  the  school’s  mission.  

0.70   0.05   0.75   0.04  

2. Everything  we  do  in  our  school  is  guided  and  directed  by  our  mission.  

0.78   0.04   0.85   0.03  

3. Our  school  mission  clearly  expresses  a  commitment  to  Catholic  identity.  

0.70   0.05   0.72   0.04  

4. Our  school  provides  an  academically  rigorous  Catholic  religion  program,  taught  by  qualified  teachers.  

0.75   0.05   0.81   0.04  

5. In  all  subjects,  teachers  help  students  think  critically  and  ethically  about  the  world  around  them,  using  the  lens  of  Gospel  values  and  Catholic  doctrine  and  

0.69   0.05   0.73   0.05  

Page 72: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  72      

beliefs.  6. Our  school  provides  opportunities  outside  

the  classroom  for  student  faith  formation,  and  participation  in  retreats,  prayer,  mass,  sacraments,  and  other  spiritual  experiences.    

0.68   0.05   0.74   0.04  

7. Our  school  provides  opportunities  outside  the  classroom  for  students  to  participate  in  service  activities  for  social  justice.    

0.72   0.05   0.72   0.05  

8. Administrators,  faculty,  and  staff  serve  as  role  models  of  faith  and  service  to  students.  

0.77   0.04   0.83   0.04  

9. Our  school  provides  opportunities  for  faith  formation  for  faculty  and  staff.  

0.83   0.04   0.88   0.03  

10. Our  school  provides  opportunities  for  faith  formation  for  parents/guardians  and  other  adult  members  of  the  school  community.  

0.85   0.03   0.82   0.03  

11. Our  school  provides  opportunities  for  adult  members  of  the  school  community  to  participate  in  service  activities  for  social  justice.  

0.68   0.05   0.75   0.05  

12. Our  school  helps  parents/guardians  support  the  faith  life  of  their  child.  

0.79   0.05   0.85   0.04  

13. Every  adult  in  the  school  supports  the  faith  life  of  the  school  community.  

0.76   0.05   0.78   0.04  

Governance  and  Leadership          14. There  is  a  person  or  group  (such  as  a  

board  or  a  pastor)  who  collaborates  with  the  school  administration  to  make  or  recommend  decisions  for  the  success  of  the  school.  

0.77   0.05   0.79   0.05  

15. Our  school  administration  takes  responsibility  for  development  and  oversight  of  school  programs,  personnel,  and  school  operations.  

0.88   0.03   0.91   0.03  

16. Our  school  administration  involves  all  members  of  the  school  community  to  make  sure  that  everything  we  do  embodies  the  school’s  mission  and  vision.  

0.84   0.03   0.87   0.03  

17. Our  school  administration  finds  ways  to  help  faculty,  staff,  students,  and  parents  collaborate  in  striving  for  excellence  in  all  aspects  of  the  school.  

0.85   0.03   0.89   0.03  

Academic  Excellence          

18. Our  school  has  a  challenging  curriculum  infused  with  Gospel  values,  preparing  students  for  college,  work,  and  life.  

0.77   0.04   0.83   0.04  

19. Curriculum  and  instruction  in  our  school  prepares  students  to  be  capable  and  critical  users  of  media  and  technology.  

0.74   0.04   0.76   0.05  

Page 73: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  73      

20. In  our  classes,  students  spend  most  of  the  time  solving  problems,  discussing  ideas,  creating  their  own  work,  reading,  writing,  speaking,  and  researching.  

0.68   0.05   0.73   0.05  

21. Teachers  use  effective  instruction  to  deliver  the  curriculum.  

0.78   0.04   0.80   0.04  

22. At  our  school,  teachers  use  different  teaching  approaches  to  meet  the  diverse  needs  of  all  students.  

0.77   0.04   0.80   0.04  

23. At  our  school,  teachers  work  together  to  help  each  other  become  better  teachers.  

0.84   0.03   0.88   0.03  

24. At  our  school,  all  administrators,  faculty  and  staff  engage  in  ongoing  professional  development.  

0.88   0.03   0.88   0.03  

25. At  our  school,  teachers  follow  school-­‐wide  policies  and  procedures  to  fairly  evaluate  and  communicate  student  performance.  

0.85   0.04   0.87   0.03  

26. At  our  school,  teachers  work  together  to  use  data  on  student  performance  to  review  and  improve  courses  and  instruction.  

0.87   0.03   0.87   0.03  

27. Teachers  vary  the  types  of  assessments  they  use  to  monitor  individual  and  class-­‐wide  student  learning.  

0.81   0.04   0.83   0.04  

28. Our  school  communicates  how  well  students  are  achieving  in  comparison  to  similar  students  locally  and/or  nationally.  

0.88   0.03   0.90   0.02  

29. Our  school  provides  programs  and  services  that  help  students  successfully  complete  the  school  program  (for  example,  guidance  and  resource  programs).  

0.81   0.04   0.83   0.04  

30. Our  school  provides  co-­‐curricular  and  extra-­‐curricular  programs  for  students  to  develop  gifts  and  talents  and  enhance  their  creative,  artistic,  social/emotional,  physical,  and  spiritual  potential.  

0.80   0.04   0.83   0.04  

31. Our  school  provides  opportunities  for  parents/guardians  to  be  involved  in  the  education  of  their  students.  

0.73   0.04   0.74   0.04  

Operational  Vitality          32. Our  school  treats  everyone  who  works  at  

the  school  with  consistency,  fairness,  and  justice.  

0.68   0.06   0.73   0.06  

33. Our  school  manages  facilities,  equipment,  and  technology  in  ways  that  enhance  teaching  and  learning.  

0.74   0.04   0.83   0.05  

34. Our  school  uses  different  ways  to  communicate  all  that  is  happening  in  our  school  to  parents/guardians,  the  school  community  and  beyond.  

0.86   0.03   0.93   0.03  

35. Our  school  does  a  good  job  of  attracting   0.71   0.05   0.81   0.04  

Page 74: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  74      

new  students  to  our  school.  

Page 75: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  75      

 

Appendix  B:  Tables  from  the  Item-­level  Analyses  of  the  Catholic  School  –  Defining  Characteristics  Surveys  

Table  B.3.  Item  Response  Frequencies  and  Intraclass  Correlation  Coefficients  (ICC)  for  the  Catholic  School  Defining  Characteristics  Survey  of  Adults  

Item  ICC  

Strongly  Disagree  

     Strongly  Agree  

Don't  Know  

Skipped  

1. Students  in  our  school  are  encouraged,  through  all  aspects  of  their  school  experience,  to  develop  a  closer  relationship  with  Jesus  Christ.  

0.07   0.8%   1.9%   6.5%   22.8%   68.0%   0.7%   3.4%  

2. Our  school  is  a  community  that  prays  together.  

0.09   1.0%   1.3%   4.4%   18.5%   74.8%   0.3%   3.6%  

3. Our  school  is  a  community  that  lives  the  Gospel  message  through  service  to  the  poor  and  to  those  in  need.  

0.08   1.2%   3.0%   7.9%   25.7%   62.3%   0.8%   3.7%  

4. Our  school  makes  Jesus  and  the  teachings  of  the  Catholic  Church  known  to  all  students.  

0.06   0.7%   1.0%   3.0%   18.3%   77.1%   0.5%   3.5%  

5. Symbols  of  the  Catholic  faith  are  displayed  throughout  our  school.  

0.07   0.6%   0.6%   1.7%   14.2%   82.9%   0.3%   3.8%  

6. Our  school  upholds  high  standards  of  excellence  in  all  it  offers.  

0.08   2.7%   4.2%   8.6%   25.2%   59.3%   0.6%   3.6%  

7. In  addition  to  academics  and  faith  formation,  our  school  offers  experiences  in  the  arts,  athletics,  and  other  extracurricular  and  service  opportunities  that  contribute  to  the  education  of  the  whole  child.  

0.11   2.0%   3.3%   8.1%   23.5%   63.2%   0.2%   3.9%  

8. Our  school  supports  the  social,  emotional  and  spiritual  growth  of  every  student.  

0.07   3.0%   3.2%   9.6%   25.5%   58.7%   0.8%   3.9%  

9. The  program  of  instruction  in  our  school  leads  students  to  seek  wisdom  and  truth,  with  a  clear  understanding  of  right  and  wrong.  

0.07   1.3%   2.6%   8.3%   29.4%   58.4%   0.9%   4.0%  

10. The  learning  environment  in  our  school  fosters  self-­‐discipline  so  that  students  can  become  more  independent  learners.  

0.07   2.2%   3.6%   11.4%   31.2%   51.5%   1.0%   4.0%  

11. Our  school  instills  in  students  the  responsibility  to  promote  Gospel  values  and  social  justice  in  the  world.  

0.07   1.1%   3.3%   9.3%   30.4%   56.0%   1.3%   4.2%  

12. Administrators  in  our  school  understand,  accept  and  model  the  teachings  of  the  Catholic  Church.  

0.10   2.0%   3.1%   7.1%   23.5%   64.3%   2.0%   4.2%  

13. The  teachers  in  our  school  understand,  promote,  demonstrate  and  teach  Catholic  values  and  beliefs.  

0.05   1.9%   2.4%   8.3%   27.8%   59.6%   1.3%   4.4%  

Page 76: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  76      

14. Our  school  helps  parents/guardians  fulfill  their  role  as  the  primary  teachers  of  the  faith  to  their  children.  

0.05   1.6%   3.6%   14.8%   34.4%   45.5%   2.1%   5.1%  

15. Everyone  connected  with  our  school  works  together,  respecting  each  other’s  gifts,  for  the  sake  of  building  a  strong,  faith-­‐filled  learning  community.  

0.07   4.1%   5.0%   13.8%   31.4%   45.6%   2.4%   5.1%  

16. Our  school  does  everything  it  can  to  eliminate  obstacles  that  hinder  or  exclude  students  from  receiving  a  Catholic  education.  

0.06   2.8%   3.8%   11.7%   26.6%   55.0%   5.6%   5.5%  

17. Our  school  operates  with  the  expressed  approval  and  support  of  our  Bishop.  

0.13   1.4%   1.1%   5.0%   19.9%   72.6%   13.5%   5.4%  

 

Table  B.4.  School-­level  Statistics  for  the  Catholic  School  Defining  Characteristics  Survey  of  Adults  

Item  

Average  Rating  

Standard  Deviation  

Minimum    Rating  

Maximum  Rating  

25th  Percentile  

Median  75th  

Percentile  

1. Students  in  our  school  are  encouraged,  through  all  aspects  of  their  school  experience,  to  develop  a  closer  relationship  with  Jesus  Christ.  

4.56   0.24   3.88   5.00   4.42   4.56   4.71  

2. Our  school  is  a  community  that  prays  together.  

4.66   0.22   4.04   5.00   4.57   4.69   4.80  

3. Our  school  is  a  community  that  lives  the  Gospel  message  through  service  to  the  poor  and  to  those  in  need.  

4.45   0.26   3.55   5.00   4.35   4.50   4.62  

4. Our  school  makes  Jesus  and  the  teachings  of  the  Catholic  Church  known  to  all  students.  

4.69   0.18   4.24   5.00   4.56   4.71   4.80  

5. Symbols  of  the  Catholic  faith  are  displayed  throughout  our  school.  

4.77   0.16   4.29   5.00   4.69   4.78   4.87  

6. Our  school  upholds  high  standards  of  excellence  in  all  it  offers.  

4.30   0.36   3.00   5.00   4.15   4.33   4.50  

7. In  addition  to  academics  and  faith  formation,  our  school  offers  experiences  in  the  arts,  athletics,  and  other  extracurricular  and  service  opportunities  that  contribute  to  the  education  of  the  whole  child.  

4.38   0.35   3.29   4.89   4.27   4.48   4.62  

8. Our  school  supports  the  social,  emotional  and  spiritual  growth  of  every  student.  

4.30   0.33   3.00   4.82   4.14   4.31   4.54  

9. The  program  of  instruction  in  our  school  leads  students  to  seek  wisdom  and  truth,  with  a  clear  understanding  of  right  and  wrong.  

4.39   0.33   2.86   5.00   4.28   4.43   4.57  

10. The  learning  environment  in  our  school  fosters  self-­‐discipline  so  that  students  can  

4.22   0.34   2.86   4.76   4.08   4.27   4.43  

Page 77: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  77      

become  more  independent  learners.  11. Our  school  instills  in  students  the  

responsibility  to  promote  Gospel  values  and  social  justice  in  the  world.  

4.36   0.26   3.64   4.88   4.20   4.37   4.50  

12. Administrators  in  our  school  understand,  accept  and  model  the  teachings  of  the  Catholic  Church.  

4.43   0.42   2.50   5.00   4.31   4.49   4.65  

13. The  teachers  in  our  school  understand,  promote,  demonstrate  and  teach  Catholic  values  and  beliefs.  

4.38   0.34   2.71   5.00   4.29   4.42   4.57  

14. Our  school  helps  parents/guardians  fulfill  their  role  as  the  primary  teachers  of  the  faith  to  their  children.  

4.19   0.26   3.57   4.69   4.01   4.21   4.39  

15. Everyone  connected  with  our  school  works  together,  respecting  each  other’s  gifts,  for  the  sake  of  building  a  strong,  faith-­‐filled  learning  community.  

4.08   0.38   2.74   4.67   3.88   4.13   4.29  

16. Our  school  does  everything  it  can  to  eliminate  obstacles  that  hinder  or  exclude  students  from  receiving  a  Catholic  education.  

4.27   0.34   3.16   5.00   4.06   4.30   4.48  

17. Our  school  operates  with  the  expressed  approval  and  support  of  our  Bishop.  

4.59   0.38   2.83   5.00   4.45   4.66   4.85  

 

Table  B.3.  Standardized  Factor  Loadings  for  the  1-­factor  CFA  Model  (School-­level)  of  the  Catholic  School  Defining  Characteristics  Survey  of  Adults  

1-­‐Factor  Model  

Item  Standardized  Factor  Loading  

Standard  Error  

1. Students  in  our  school  are  encouraged,  through  all  aspects  of  their  school  experience,  to  develop  a  closer  relationship  with  Jesus  Christ.  

0.90   0.04  

2. Our  school  is  a  community  that  prays  together.  

0.85   0.06  

3. Our  school  is  a  community  that  lives  the  Gospel  message  through  service  to  the  poor  and  to  those  in  need.  

0.74   0.08  

4. Our  school  makes  Jesus  and  the  teachings  of  the  Catholic  Church  known  to  all  students.  

0.96   0.04  

5. Symbols  of  the  Catholic  faith  are  displayed  throughout  our  school.  

0.89   0.06  

6. Our  school  upholds  high  standards  of  excellence  in  all  it  offers.  

0.88   0.04  

7. In  addition  to  academics  and  faith  formation,  our  school  offers  experiences  in  the  arts,  athletics,  and  other  extracurricular  and  service  opportunities  

0.66   0.10  

Page 78: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  78      

that  contribute  to  the  education  of  the  whole  child.  

8. Our  school  supports  the  social,  emotional  and  spiritual  growth  of  every  student.  

0.93   0.03  

9. The  program  of  instruction  in  our  school  leads  students  to  seek  wisdom  and  truth,  with  a  clear  understanding  of  right  and  wrong.  

0.98   0.02  

10. The  learning  environment  in  our  school  fosters  self-­‐discipline  so  that  students  can  become  more  independent  learners.  

0.85   0.05  

11. Our  school  instills  in  students  the  responsibility  to  promote  Gospel  values  and  social  justice  in  the  world.  

0.94   0.04  

12. Administrators  in  our  school  understand,  accept  and  model  the  teachings  of  the  Catholic  Church.  

0.94   0.03  

13. The  teachers  in  our  school  understand,  promote,  demonstrate  and  teach  Catholic  values  and  beliefs.  

0.95   0.04  

14. Our  school  helps  parents/guardians  fulfill  their  role  as  the  primary  teachers  of  the  faith  to  their  children.  

0.93   0.05  

15. Everyone  connected  with  our  school  works  together,  respecting  each  other’s  gifts,  for  the  sake  of  building  a  strong,  faith-­‐filled  learning  community.  

0.91   0.04  

16. Our  school  does  everything  it  can  to  eliminate  obstacles  that  hinder  or  exclude  students  from  receiving  a  Catholic  education.  

0.92   0.05  

17. Our  school  operates  with  the  expressed  approval  and  support  of  our  Bishop.  

0.72   0.10  

 

Table  B.4.  Item  Response  Frequencies  and  Intraclass  Correlation  Coefficients  (ICC)  for  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (5-­8  Grades)  

Item  ICC  

Strongly  Disagree  

     Strongly  Agree  

Don't  Know  

Skipped  

1. My  school  teaches  me  to  love  Jesus.   0.09   0.8%   0.9%   3.9%   16.3%   78.1%   0.5%   1.6%  2. In  my  school,  we  pray  throughout  the  

day.  0.11   1.6%   2.6%   7.9%   19.4%   68.5%   0.3%   1.7%  

3. We  do  things  at  my  school  to  serve  the  poor  and  those  in  need.  

0.07   2.0%   3.6%   13.7%   30.4%   50.4%   1.7%   2.1%  

4. All  students  in  my  school  learn  about  Jesus  and  what  Catholics  believe.  

0.06   1.3%   1.5%   4.8%   19.5%   72.9%   1.3%   2.3%  

5. There  are  crucifixes  and  other  Catholic  symbols  throughout  the  school.  

0.09   1.0%   1.2%   3.3%   13.0%   81.5%   1.5%   2.0%  

6. My  school  strives  to  be  excellent  in  all  areas.  

0.06   4.7%   5.0%   13.9%   29.8%   46.6%   3.0%   2.2%  

Page 79: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  79      

7. My  school  offers  extra-­‐curricular  activities  and  service  projects.  

0.09   3.2%   4.7%   10.3%   27.2%   54.6%   2.6%   2.4%  

8. Adults  in  my  school  show  that  they  care  about  how  students  are  doing.  

0.08   5.8%   6.8%   14.1%   29.7%   43.6%   3.3%   1.9%  

9. It  is  important  at  my  school  to  learn  what  is  right  and  what  is  wrong,  and  act  accordingly.  

0.04   1.9%   2.0%   6.8%   24.9%   64.4%   1.2%   2.2%  

10. Students  at  my  school  know  how  to  behave  and  respect  others  so  that  everyone  can  learn.  

0.05   13.8%   13.6%   24.6%   26.9%   21.1%   1.4%   2.1%  

11. Even  during  subjects  other  than  religion,  we  learn  to  act  according  to  what  Jesus  taught,  and  to  help  make  the  world  a  better  place  for  everyone.  

0.08   6.4%   9.2%   19.2%   31.0%   34.2%   1.6%   2.7%  

12. The  principal  leads  us  in  prayer  and  is  involved  in  school  Masses,  and  makes  sure  everyone  knows  that  my  school  is  a  Catholic  school.  

0.18   5.0%   5.2%   10.4%   22.5%   57.0%   1.8%   2.5%  

13. The  teachers  in  my  school  teach  and  act  in  a  way  that  shows  they  understand  Catholic  values  and  beliefs.  

0.06   4.0%   4.7%   12.0%   30.3%   49.0%   1.6%   2.6%  

14. The  teachers  and  the  principal  at  my  school  show  respect  for  parents/guardians,  and  keep  them  well  informed  about  their  students  and  the  school.  

0.08   4.5%   5.0%   11.4%   28.6%   50.4%   1.8%   2.6%  

15. My  school  works  together  in  ways  that  show  respect  for  everyone’s  unique  talents.  

0.06   7.3%   7.6%   16.6%   31.3%   37.2%   2.0%   2.6%  

16. My  school  is  a  community  that  welcomes  all  students.  

0.09   4.3%   4.3%   8.9%   22.0%   60.5%   2.1%   3.0%  

17. The  students  in  my  school  know  that  the  school  is  part  of  the  diocese  and  connected  to  the  Bishop.  

0.06   4.1%   4.5%   11.8%   27.8%   51.7%   9.4%   3.1%  

 

Table  B.5.  School-­level  Statistics  for  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (5-­8  Grades)  

Item  

Average  Rating  

Standard  Deviation  

Minimum    Rating  

Maximum  Rating  

25th  Percentile  

Median  75th  

Percentile  

1. My  school  teaches  me  to  love  Jesus.   4.72   0.17   4.23   5.00   4.64   4.72   4.84  

2. In  my  school,  we  pray  throughout  the  day.  4.49   0.34   3.00   5.00   4.31   4.50   4.72  

3. We  do  things  at  my  school  to  serve  the  poor  and  those  in  need.  

4.24   0.31   3.50   5.00   4.12   4.27   4.40  

4. All  students  in  my  school  learn  about  Jesus  and  what  Catholics  believe.  

4.64   0.21   4.00   5.00   4.54   4.67   4.76  

5. There  are  crucifixes  and  other  Catholic  symbols  throughout  the  school.  

4.73   0.19   4.18   5.00   4.64   4.78   4.88  

Page 80: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  80      

6. My  school  strives  to  be  excellent  in  all  areas.  

4.15   0.38   3.00   5.00   3.97   4.21   4.37  

7. My  school  offers  extra-­‐curricular  activities  and  service  projects.  

4.24   0.41   2.50   4.83   4.07   4.31   4.48  

8. Adults  in  my  school  show  that  they  care  about  how  students  are  doing.  

4.03   0.53   1.50   5.00   3.84   4.01   4.24  

9. It  is  important  at  my  school  to  learn  what  is  right  and  what  is  wrong,  and  act  accordingly.  

4.54   0.21   4.04   5.00   4.46   4.53   4.67  

10. Students  at  my  school  know  how  to  behave  and  respect  others  so  that  everyone  can  learn.  

3.35   0.46   2.00   4.50   3.04   3.31   3.63  

11. Even  during  subjects  other  than  religion,  we  learn  to  act  according  to  what  Jesus  taught,  and  to  help  make  the  world  a  better  place  for  everyone.  

3.90   0.41   3.12   5.00   3.55   3.80   4.20  

12. The  principal  leads  us  in  prayer  and  is  involved  in  school  Masses,  and  makes  sure  everyone  knows  that  my  school  is  a  Catholic  school.  

4.25   0.55   2.00   4.89   4.05   4.42   4.65  

13. The  teachers  in  my  school  teach  and  act  in  a  way  that  shows  they  understand  Catholic  values  and  beliefs.  

4.20   0.38   3.00   5.00   4.01   4.21   4.42  

14. The  teachers  and  the  principal  at  my  school  show  respect  for  parents/guardians,  and  keep  them  well  informed  about  their  students  and  the  school.  

4.20   0.37   3.00   4.80   4.07   4.22   4.42  

15. My  school  works  together  in  ways  that  show  respect  for  everyone’s  unique  talents.  

3.88   0.44   2.00   4.48   3.73   3.88   4.23  

16. My  school  is  a  community  that  welcomes  all  students.  

4.38   0.31   3.44   5.00   4.18   4.40   4.63  

17. The  students  in  my  school  know  that  the  school  is  part  of  the  diocese  and  connected  to  the  Bishop.  

4.20   0.57   1.00   4.83   4.10   4.28   4.46  

 

Page 81: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  81      

 

Table  B.6.  Standardized  Factor  Loadings  for  the  1-­factor  CFA  Model  (School-­level)  of  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (5-­8  Grades)  

1-­‐Factor  Model  

Item  Standardized  Factor  Loading  

Standard  Error  

1. My  school  teaches  me  to  love  Jesus.   0.88   0.06  2. In  my  school,  we  pray  throughout  the  

day.  0.47   0.15  

3. We  do  things  at  my  school  to  serve  the  poor  and  those  in  need.  

0.47   0.14  

4. All  students  in  my  school  learn  about  Jesus  and  what  Catholics  believe.  

0.85   0.06  

5. There  are  crucifixes  and  other  Catholic  symbols  throughout  the  school.  

0.70   0.10  

6. My  school  strives  to  be  excellent  in  all  areas.  

0.91   0.05  

7. My  school  offers  extra-­‐curricular  activities  and  service  projects.  

0.62   0.11  

8. Adults  in  my  school  show  that  they  care  about  how  students  are  doing.  

0.96   0.04  

9. It  is  important  at  my  school  to  learn  what  is  right  and  what  is  wrong,  and  act  accordingly.  

0.96   0.04  

10. Students  at  my  school  know  how  to  behave  and  respect  others  so  that  everyone  can  learn.  

0.74   0.10  

11. Even  during  subjects  other  than  religion,  we  learn  to  act  according  to  what  Jesus  taught,  and  to  help  make  the  world  a  better  place  for  everyone.  

0.84   0.07  

12. The  principal  leads  us  in  prayer  and  is  involved  in  school  Masses,  and  makes  sure  everyone  knows  that  my  school  is  a  Catholic  school.  

0.78   0.08  

13. The  teachers  in  my  school  teach  and  act  in  a  way  that  shows  they  understand  Catholic  values  and  beliefs.  

0.95   0.04  

14. The  teachers  and  the  principal  at  my  school  show  respect  for  parents/guardians,  and  keep  them  well  informed  about  their  students  and  the  school.  

0.86   0.05  

15. My  school  works  together  in  ways  that  show  respect  for  everyone’s  unique  talents.  

0.97   0.03  

16. My  school  is  a  community  that  welcomes  all  students.  

0.92   0.04  

17. The  students  in  my  school  know  that  the   0.85   0.07  

Page 82: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  82      

school  is  part  of  the  diocese  and  connected  to  the  Bishop.  

 

Page 83: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  83      

 

Table  B.7.  Item  Response  Frequencies  for  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (9-­12  Grades)  

Item  

Strongly  Disagree  

     Strongly  Agree  

Don't  Know  

Skipped  

1. Students  in  our  school  are  encouraged,  through  all  aspects  of  their  school  experience,  to  develop  a  closer  relationship  with  Jesus  Christ.  

2.5%   3.3%   10.7%   24.0%   59.5%   0.0%   5.5%  

2. Our  school  is  a  community  that  prays  together.  

2.5%   0.8%   2.5%   23.1%   71.1%   0.0%   5.5%  

3. Our  school  is  a  community  that  lives  the  Gospel  message  through  service  to  the  poor  and  to  those  in  need.  

1.7%   1.7%   12.4%   31.4%   52.9%   0.0%   5.5%  

4. Our  school  makes  Jesus  and  the  teachings  of  the  Catholic  Church  known  to  all  students.  

0.8%   0.8%   5.8%   26.4%   66.1%   0.0%   5.5%  

5. Symbols  of  the  Catholic  faith  are  displayed  throughout  our  school.  

0.8%   0.8%   7.6%   31.4%   59.3%   0.8%   7.0%  

6. Our  school  upholds  high  standards  of  excellence  in  all  it  offers.  

1.7%   2.5%   9.2%   34.5%   52.1%   0.8%   6.3%  

7. In  addition  to  academics  and  faith  formation,  our  school  offers  experiences  in  the  arts,  athletics,  and  other  extracurricular  and  service  opportunities  that  contribute  to  the  education  of  the  whole  child.  

0.8%   0.8%   8.3%   24.8%   65.3%   0.0%   5.5%  

8. Our  school  supports  the  social,  emotional  and  spiritual  growth  of  every  student.  

2.5%   1.7%   9.2%   35.3%   51.3%   0.0%   7.0%  

9. The  program  of  instruction  in  our  school  leads  students  to  seek  wisdom  and  truth,  with  a  clear  understanding  of  right  and  wrong.  

1.7%   0.8%   9.9%   38.0%   49.6%   0.0%   5.5%  

10. The  learning  environment  in  our  school  fosters  self-­‐discipline  so  that  students  can  become  more  independent  learners.  

1.7%   3.3%   14.0%   30.6%   50.4%   0.0%   5.5%  

11. Our  school  instills  in  students  the  responsibility  to  promote  Gospel  values  and  social  justice  in  the  world.  

3.3%   1.7%   12.4%   38.0%   44.6%   0.0%   5.5%  

12. Administrators  in  our  school  understand,  accept  and  model  the  teachings  of  the  Catholic  Church.  

0.8%   1.7%   9.9%   37.2%   50.4%   0.0%   5.5%  

13. The  teachers  in  our  school  understand,  promote,  demonstrate  and  teach  Catholic  values  and  beliefs.  

0.8%   0.8%   16.5%   31.4%   50.4%   0.0%   5.5%  

14. The  adults  in  our  school  show  that  they  are  partners  with  our  parents/guardians  in  our  Catholic  education.  

1.8%   1.8%   15.8%   30.7%   50.0%   3.9%   7.0%  

Page 84: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  84      

15. Everyone  connected  with  our  school  works  together,  respecting  each  other’s  gifts,  for  the  sake  of  building  a  strong,  faith-­‐filled  learning  community.  

4.3%   2.6%   12.1%   36.2%   44.8%   2.3%   7.0%  

16. Our  school  does  everything  it  can  to  eliminate  obstacles  that  hinder  or  exclude  students  from  receiving  a  Catholic  education.  

2.6%   1.8%   13.2%   28.9%   53.5%   3.1%   7.8%  

17. Our  school  operates  with  the  expressed  approval  and  support  of  our  Bishop.  

2.7%   0.0%   9.8%   24.1%   63.4%   5.5%   7.0%  

 

Table  B.8.  Standardized  Factor  Loadings  for  the  1-­factor  CFA  Model  (Student-­level)  of  the  Catholic  School  Defining  Characteristics  Survey  of  Students  (9-­12  Grades)  

1-­‐Factor  Model  

Item  Standardized  Factor  Loading  

Standard  Error  

1. Students  in  our  school  are  encouraged,  through  all  aspects  of  their  school  experience,  to  develop  a  closer  relationship  with  Jesus  Christ.  

0.86   0.03  

2. Our  school  is  a  community  that  prays  together.  

0.87   0.03  

3. Our  school  is  a  community  that  lives  the  Gospel  message  through  service  to  the  poor  and  to  those  in  need.  

0.88   0.03  

4. Our  school  makes  Jesus  and  the  teachings  of  the  Catholic  Church  known  to  all  students.  

0.91   0.03  

5. Symbols  of  the  Catholic  faith  are  displayed  throughout  our  school.  

0.85   0.03  

6. Our  school  upholds  high  standards  of  excellence  in  all  it  offers.  

0.89   0.03  

7. In  addition  to  academics  and  faith  formation,  our  school  offers  experiences  in  the  arts,  athletics,  and  other  extracurricular  and  service  opportunities  that  contribute  to  the  education  of  the  whole  child.  

0.82   0.04  

8. Our  school  supports  the  social,  emotional  and  spiritual  growth  of  every  student.  

0.83   0.03  

9. The  program  of  instruction  in  our  school  leads  students  to  seek  wisdom  and  truth,  with  a  clear  understanding  of  right  and  wrong.  

0.85   0.03  

10. The  learning  environment  in  our  school  fosters  self-­‐discipline  so  that  students  can  become  more  independent  learners.  

0.88   0.03  

Page 85: Survey Validity and reliability Study · PDF filevalidity%and%reliability%of ... items%can%be%problematic%as%high%rates%of%missing%datanegatively%affectthe%internal%validity%and% generalizability%of%the%survey

Page  |  85      

11. Our  school  instills  in  students  the  responsibility  to  promote  Gospel  values  and  social  justice  in  the  world.  

0.89   0.02  

12. Administrators  in  our  school  understand,  accept  and  model  the  teachings  of  the  Catholic  Church.  

0.78   0.03  

13. The  teachers  in  our  school  understand,  promote,  demonstrate  and  teach  Catholic  values  and  beliefs.  

0.82   0.03  

14. The  adults  in  our  school  show  that  they  are  partners  with  our  parents/guardians  in  our  Catholic  education.  

0.83   0.03  

15. Everyone  connected  with  our  school  works  together,  respecting  each  other’s  gifts,  for  the  sake  of  building  a  strong,  faith-­‐filled  learning  community.  

0.86   0.03  

16. Our  school  does  everything  it  can  to  eliminate  obstacles  that  hinder  or  exclude  students  from  receiving  a  Catholic  education.  

0.91   0.03  

17. Our  school  operates  with  the  expressed  approval  and  support  of  our  Bishop.  

0.75   0.05