using the idb analyzer to analyze iccs data
DESCRIPTION
Using the IDB Analyzer to Analyze ICCS data. Table of content. I. Overview Starting the Analysis Module Overview of Analysis Methods II. Percentages only III. Percentages and Means IV. Correlations V. Regression VI. Benchmarks VII. Hands-On Training. Starting the Analysis Module. - PowerPoint PPT PresentationTRANSCRIPT
ICCS 2009 IDB Seminar – Nov 24-26, 2010 – IEA DPC, Hamburg, Germany
Using the IDB Analyzerto Analyze ICCS data
Table of contentI. Overview
– Starting the Analysis Module– Overview of Analysis Methods
II. Percentages onlyIII. Percentages and MeansIV. CorrelationsV. RegressionVI. BenchmarksVII. Hands-On Training
To start the Analysis Module of the IDB Analyzer, select:Start > Programs > IEA > IDB Analyzer > Analysis Module
Starting the Analysis Module
Overview of the Analysis Module
To start the IDB Analyzer Analysis Module choose:Start > Programs > IEA > IDB Analyzer > Analysis Module
Overview of the Analysis Module
Choose file containing the data for analysis
Overview of the Analysis Module
Overview of the Analysis Module
Percentages and Means– Computes means and standard deviation on continuous
variables for specified subgroups, displays percentages of cases within those subgroups
– Also computes the appropriate standard errors for those percentages, means and standard deviations
Overview of the Analysis Module
Percentages Only– Percentages of variables with their BRR standard errors– Computes the percentages of participants within specified
subgroups– Computes the appropriate standard errors for those
percentages
Overview of the Analysis Module
Regression– Calculates a multiple linear regression between a dependent
variable and a set of independent variables– Computes the regression coefficients and their
corresponding standard errors– Can be used to compare means of subgroups and their
significance
Overview of the Analysis Module
Correlations– Calculates correlation coefficients for selected analysis
variables and their BRR standard errors
11
IEA IDB Analyzer – Analysis Module
Benchmarks– Computes percentages of students within, reaching or
surpassing user provided benchmarks (Proficiency Level) of achievement with JRR standard errors for those percentages
Overview of the Analysis Module
Clear all selections: clears all the settings defined by the user– Selected analysis type– Selected grouping, analysis, dependent, weight and replicate
weight variables– Defined output files
Overview of the Analysis Module
Select study variables of interest
Search for variable by name or label
Overview of the Analysis Module
Add or remove variables
Overview of the Analysis Module
Applies to all analysis types
Overview of the Analysis Module
Applies to:- Percentages and Means- Regression- Correlations
Overview of the Analysis Module
Applies to:- Regression
Overview of the Analysis Module
Selected automatically when file is opened
Overview of the Analysis Module
Choose the desired number of decimals in the output
Overview of the Analysis Module
Define the location of the output files to be produced
Table of contentI. Overview
– Starting the Analysis Module– Overview of Analysis Methods
II. Percentages onlyIII. Percentages and MeansIV. CorrelationsV. RegressionVI. BenchmarksVII. Hands-On Training
Student Questionnaire, Q35, p. 34Variable of Interest: IS2P35
Attendance of religious servicesHow frequently students attend religious services outside home together with other people?
Reproducing Table 4.12 from ICCS2009 International report, the last five columnsAnalysis File: merged in the previous step School and Student Background filesC:\ICCS2009\Work\ICG_ISG_INTC2.savAnalysis Type: “Percentages only” with “Exclude Missing from Analysis” checkedGrouping Variables:IS2P35 – Frequency of religious practices outside home[IDCNTRY] – Country IDs (Pre-selected)
Percentages Only
Percentages Only - Settings
C:\ICCS2009\Work\Table_4-12.*
C:\ICCS2009\Work\ICG_ISG_INTC2.sav
IDCNTRY IS2P35
TOTWGTS
JKZONES1
Percentages onlyExclude Missing from Analysis
Percentages Only
The IDB Analyzer creates SPSS Syntax and starts SPSSIn SPSS Syntax Editor Choose: Run > All
Percentages Only
As a result the IDB Analyzer creates the following in the working directory (C:\ICCS2009\Work\):
SPSS Syntax file – contains the syntax with the commands (*.sps)SPSS Data file – contains statistics from the analysis (*.sav)MS Excel Output file – contains statistics from the analysis (*.xls)
Percentages Only - SPSS Output
Number of students in each group
List of countries
Variable name and Value Labels
Percentages Only - Excel Output
Percentages Only - Excel Output
Total student weight – population estimate of the groups defined by the grouping variables(IDCNTRY, IS2P35)
Weighted percentage of students for the groups defined by the grouping variables (IDCNTRY, IS2P35)
Standard error of percentages
Percentages Only - Excel Output
Percentages Only - Interpretation
On the average 28% of the students across countries never attend religious services outside home with others. This percentage is highest in Czech republic (70%) and lowest in Indonesia (6.5%).
Table of contentI. Overview
– Starting the Analysis Module– Overview of Analysis Methods
II. Percentages onlyIII. Percentages and MeansIV. CorrelationsV. RegressionVI. BenchmarksVII. Hands-On Training
Student Questionnaire, Q02, p. 2Variable of Interest: SGENDER
Distribution of civic knowledge by gender
What is the distribution of students’ civic knowledge in regard with their gender?
Reproducing Table 3.13 from ICCS2009 International report, the last five columns
Percentages & Means - Settings
C:\ICCS2009\Work\Table_3-13.*
C:\ICCS2009\Work\ICG_ISG_INTC2.sav
TOTWGTS
JKZONES
IDCNTRY SGENDER
PVCIV01-05
1
Percentages and MeansExclude Missing from Analysis
Percentages & Means – SPSS Output
Number of students in the sample
List of countries defined by the first groping variable (IDCNTRY)
Achievement scores(PVCIV01-05)
Percentages & Means – Excel Output
Groups of students defined by the second grouping variable (SGENDER)
Weighted percentages of students for the groups defined by the grouping variables (IDCNTRY, SGENDER)
Standard error of percentages
Percentages & Means – Excel Output
Total student weight – population estimate of the group defined by the grouping variable(IDCNTRY)
Mean of the analysis variable (PV5CIV01-05)
Standard error of the mean of the analysis variable
Standard deviation of the analysis variable
Standard error of the standard deviation
Percentages & Means – Excel Output
Percentages & Means – Interpretation
In all countries girls outperform boys in regard with civic knowledge. These differences are lowest in Guatemala and greatest in Thailand.
Table of contentI. Overview
– Starting the Analysis Module– Overview of Analysis Methods
II. Percentages onlyIII. Percentages and MeansIV. CorrelationsV. RegressionVI. BenchmarksVII. Hands-On Training
Correlation (Pearson)
Measures the linear relationship between two (continuous) variablesCorrelation coefficient r informs us about the strength and direction of this relationshipScatter plot can give you a first idea if there is a connection between the variables of interestCorrelation coefficient is a sample estimate of a population parameter too!If we want to know how confident we can be in that value, we need to calculate its standard error that accounts for the complex sampling design
Student Questionnaire, Q14A-F, p. 13Variables of Interest: PARTCOM, PVCIV01-05
Participation in communityIs the student participation in wider community associated with their civic knowledge?
Correlations - Settings
C:\ICCS2009\Work\Correlations.*
C:\ICCS2009\Work\ICG_ISG_INTC2.sav
TOTWGTS
JKZONES
IDCNTRY
PVCIV01-05
PARTCOM
1
CorrelationsExclude Missing from Analysis
Correlations – SPSS Output
Correlations – Excel Output
List of Countries
Achievement Scores
Sum of Weights
Mean achievement
S.E. of the mean achievement
Correlations – Excel Output
Standard Deviation of Mean Achievement
S.E. of the standard deviation of the mean achievement
Correlation of the set of PV with itself
Correlations – Excel OutputS.E. of correlation of the set of PV with itself
Mean of the analysis variable
S.E. of the mean of the analysis variable
Analysis variable
Correlations – Excel Output
Standard deviation of the analysis variable
S.E. of the correlation between the set of PVs and the analysis variable
S.E. of the standard deviation of the analysis variable
Correlation between the set of PVs and the analysis variable
Correlations – Interpretation
Most countries show very weak, negative relationship between the participation in community and civic knowledge. However, Latin American countries and Indonesia show moderately strong negative correlation (Guatemala r=-0.28)
Table of contentI. Overview
– Starting the Analysis Module– Overview of Analysis Methods
II. Percentages onlyIII. Percentages and MeansIV. CorrelationsV. RegressionVI. BenchmarksVII. Hands-On Training
Linear Regression Model
Linear Regression Model
y is the dependent variable – here: estimated mean of all 5 plausible valuesx is the independent variableß0 is the intercept (value of y when x is zero)
ß1 is the slope (change in y for each unit increase in x)
10 ββ + xy
Recoding variablesReproducing Table 7.8 from ICCS2009 International report, columns 2 and 10Most regressions in Table 7.8 has been calculated using dummy recoded variablesInformation of the students‘ immigration background (variable IMMIG) needs to be recoded to dummy variableFor replication of the table, the information from the variable IMMIG needs to be recoded in SPSS and added as a new variable to the data file before running the analysis
Native1st
Generation Immigrant
Non-Native
IMMIG 1 2 3
Reg01IMMIG 0 1 1
Dummy Coding for Regression
IMMIG REGIMMIG
Dummy Coding for Regression
IMMIG values:
0
1
1
System Missing
Menu: TRANSFORM Recode into Different Variables...
SPSS: Dummy Coding for SPSS: Dummy Coding for RegressionRegression
SPSS: Dummy Coding for SPSS: Dummy Coding for RegressionRegression
1 02 13 1ELSE SYSMISS
Menu: TRANSFORM Recode into Different Variables...
SPSS: Dummy Coding for SPSS: Dummy Coding for RegressionRegression
Save the file under different name:C:\ICCS2009\Work\ICG_ISG_INTC2_REG.sav
SPSS: Dummy Coding for SPSS: Dummy Coding for RegressionRegressionOr just run the following syntax (C:\ICCS2009\Work\Recode_IMMIG.sps):
GET FILE = "C:\ICCS2009\Work\ICG_ISG_INTC2.sav".
RECODE IMMIG (MISSING=SYSMISS) (1=0) (ELSE=1) INTO REGIMMIG.VARIABLE LABELS REGIMMIG "RECODED IMMIGRATION BACKGROUND".VALUE LABELS REGIMMIG "0" "No immigration background" "1" "Immigration Background".EXECUTE.
SAVE OUTFILE = "C:\ICCS2009\Work\ICG_ISG_INTC2_REG.sav".
10 ββ + xy
Linear Regression Model
Predictor variable: REGIMMIG
Mean achievement for native students
Difference between mean achieve-ment of native and mean achieve-ment of non-
native students
IDCNTRY
TOTWGTS
JKZONES
C:\ICCS2009\Work\ICG_ISG_INTC2_REG.sav
C:\ICCS2009\Work\Table_7-08.*
REGIMMIG
PVICIV01-05
RegressionWith Achievement Scores Exclude Missing from Analysis
Regression - Settings
Regression – SPSS Output
Number of Cases
Multiple R-Squared
Intercept – the mean achievement of native students
S.E. of Intercept
List of countries
Regression – SPSS Output
Regression coefficient for the independent variable
S.E. of regression coefficient
t-test statistics
Regression – SPSS Output
In 27 out of 38 countries there are statistically significant difference in civic knowledge between native students and non-native students.
ABS (t-test) > 1.96
↓The difference is statistically
signifficant
Regression – Interpretation
In Austria the difference in civic knowledge between native students and non-native students is statistically significant, but not in Bulgaria.
Austria – ABS (-8.89) > 1.96 Bulgaria – ABS (-1.27) < 1.96
Regression – Interpretation
Mean achievement for native students
Difference between mean achievement of native and mean
achievement of non-native students
56.6516.2 -= xy
Table of contentI. Overview
– Starting the Analysis Module– Overview of Analysis Methods
II. Percentages onlyIII. Percentages and MeansIV. CorrelationsV. RegressionVI. BenchmarksVII. Hands-On Training
Benchmarks/Proficiency Levels
In ICCS the proficiency levels are distinct areas in the distribution providing descriptions of achievement on the scale in relation to performance on the questions asked
Level 3 of International Proficiency Level: ≥ 563 score pointsLevel 2 of International Proficiency Level: 479 562 score pointsLevel 1 of International Proficiency Level: 396 478 score points
IEA IDB Analyzer computes percentages of students within, reaching or surpassing user provided proficiency levels of achievement with JRR standard errors for those percentages
Percentages of Students at Specific Proficiency Levels
What is the percentage of students reaching each one of the ICCS 2009 proficiency levels in each country?
Reproducing Table 3.12 from ICCS2009 International report
Benchmarks - Settings
C:\ICCS2009\Work\Table_3-12.*
C:\ICCS2009\Work\ICG_ISG_INTC2_REG.sav
IDCNTRY
TOTWGTS
JKZONES
563 479 395
BenchmarksExclude Missing from Analysis
1
Benchmarks - SPSS Output
Performance groups
List of countries
Achievement scores variable name
Benchmarks - Excel Output
Proficiency levels cut-points
Benchmarks - Excel Output
S.E. of the percentage
Estimated number of students
Total student weight – contains estimates in the population in each group
Weighted percentage of students below or within specific Proficiency Level
Benchmarks - Interpretation
The proportion of students in approximately half of the countries reach or excede the international average for Proficiency Level 3 (28%). The highest percentage of students reaching or exceding Level 3 is Finland, and the lowest is the Dominican Republic.
Any Questions?
Thank you for your attention!
Table of contentI. Overview
– Starting the Analysis Module– Overview of Analysis Methods
II. Percentages onlyIII. Percentages and MeansIV. CorrelationsV. RegressionVI. BenchmarksVII. Hands-On Training
Hands-On TrainingA. Re-produce the presented example using your
country data- Percentage of students’ agreement to the right of expressing opinions freely (IS2P20A)
and/orB. Re-produce the presented example using your
country data- Mean civic and civic knowledge of students by their gender
and/orC. Practice with own selected variables following these
analysis steps (Percentages and/or Percentages and Means)
Hands-On Training
A. Re-produce the example using your country data- Correlation of students’ participation in the wider community with civic and citizenship achievement (PVCIV01-05 with PARTCOM)
and/orB. Re-produce the example using your country data
- Regression of students’ immigration status on civic and citizenship achievement of (IMMIG [recoded] on PVCIV01-05)
and/orC. Practice with own selected variables following these
analysis steps (Correlation and/or Regression)