2xk

with edifficexamscheThe d ProcThereeffec Initia

2-wa

Main

Initia Get unian / m / pr / pl / de

2-Factor

The purpoexam performulty as the ex

m problems (=dule showed dependent va

cess: e are a lot of

cts will require

al Analysis Get descr Determine Consider

ay Interactio Get 2-way

n Effects Get estim Why are t

al Analysis

descriptive

nova testperf

method = sstyprint descriptivot profile (pgesign = pg1e

Between G

ose of this stumance. Practxam problems=3). Different

the class meariable was pe

steps to a coe different sub

riptive meanse what effectswhat main ef

ons y cell means &

ated marginathe “Descriptiv

e means, plo

f by ar1y2n

pe(3) es 1e2h3s*ar1y2

e2h3s ar1y2

Groups AN

udy was to exice Difficulty w

s (=1), they wsections of th

eeting during werformance o

mplete analysbsets of these

, plots & F-tess are significaffects are likel

& follow-up a

al means & fove” and “Estim

ots & F-test

pg1e2h3s

2n) 2n pg1e2h3s

NOVA

xamine the relwas a 3-cond

were easier thhe course wewhich the exan an examina

sis of a 2-waye. Here’s a pre

sts ant ly to be intere

nalyses to de

llow-up analymated” margi

ts

s*ar1y2n.

lationships ofdition variable han the exam ere randomly aam review woation.

y design. Diffeview…

esting – based

escribe the 2-w

yses to descrinal means dif

lists DV order deteStatistics

corrects e get descr get plot o specify th

automatic

The “Descr The marginthe cell mea For exampl ( (44.00 * 10

f exam Review - practice prproblems (=2assigned to re

ould occur & s

ferent pattern

d on the aggr

way interactio

be each mainfferent ?

“by” IVs ermines left-ttable

each effect foriptive cell andof cell means he design inclcally calculate

riptive Statistic

nal means areans being agg

e, the margin

0) + (61.667

w Attendanceroblems were2), or they weeceive the thr

student’s atten

ns of significa

regations invo

on

n effect

to-right orderi

or all other effed marginal me(x-axis * “sepuding the intees from the IV

cs” are the ra

e weighted bygregated.

nal mean for

* 6) ) / 16 = 5

e and Practicee either about ere more difficree difficulty lndance was r

nt and non-si

olved

ng of IVs in th

ects eans

parate lines” )eraction that iVs specified a

aw or “uncorre

y the different

the Easier Pr

50.625

e Difficulty the same

cult than the evels. The recorded.

gnificant

he Descriptive

s above)

ected” means

ial sizes of

ractDif is

e

.

Fliwrpa

From the meaike Harder anwork best for treview, but Eapractice work attend the rev

We “all

ans and the pnd Same difficthose who doasier and Sambest for those

view.

have significaround” !

lots, it looks culty practice o attend the me difficulty e who do not

ant effects

Consider what lower-order effects we will need to check for descriptive/misleading patterns Because of the significant 2-way, the means patterns of each main effect will have to be carefully checked against the corresponding simple effects to determine if they are descriptive or misleading. Remember, this will have to be done whether the main effect is significant or not – main effect nulls can be misleading! Consider what lower-order effects are likely to be interesting – based on the aggregations involved PractDif

These conditions are really pretty arbitrary. More importantly, it is unclear what population is represented by an average of those who attended and not attend

the review session! So, this main effect is only likely to be interesting if that main effect is descriptive, and so, it describes the

behavior of both those who did and did not attend the review. Attend the Review

This is a straightforward operationalization of a simple variable However, the marginal means are of dubious value, because the PractDif conditions are arbitrary, and so it is not

clear what population would be represented by the aggregate of the easier, harder, and similar difficulty performances

So, this main effect is only likely to be interesting if that main effect is descriptive, and so, it describes the behavior of those who practiced with similarly difficult, harder, and easier materials.

Remember – – non-significant lower-order effects that are involved in a significant higher order effect must be compared to the corresponding simple effects, to determine whether they are descriptive or misleading!!! 2-way Interaction Pairwise Comparisons You will usually want both sets of simple effects. One of those sets will be used to describe the pattern of the significant interaction. Each set will be used to determine if the corresponding main effect pattern is descriptive or misleading. Select the set of simple effects that most directly addresses the research question or research hypothesis The statement that, “We wanted to know if the relative difficulty of the practice material was related to test performance, and if this effect was different for those who did and did not attend the review session.” makes the selection of the simple effects to use to describe the interaction straightforward. From this, we’ll want to focus on the simple effect of practice difficulty (easier, harder, similar) and then examine how this simple effect is different those who did and did not attend the review session.

ObtaAtte / em

The p This

aining and dndance

mmeans table

pairwise effec

interaction pa

describing

es (ar1y2n by

cts describing Did atten

Did not a

attern allows

the pairwis

y pg1e2h3s )

g the interactio nd the review

attend review

us to anticipa

e simple ef

compare ( pg

on are:

w

w

ate that the ma

ffects of Pra

g1e2h3s )

Easier v Sa

<

=

ain effect patt

actice Diffic

this asksfor the 2

the orde“table” cmeans

the variacommantest

Same cell mabove, but rcommand.

The F-tessimple efcondition With 3 Prfollow-upthese sim

me Easie

tern of Practic

culty for eac

s for the an a2-way interacter of the variacommand con

able specifiednd tells which

means as in threarranged to

sts tell us thatffect of Practic of Review At

ractice Difficup analyses to emple effects.

er v Harder

<

>

ce Difficulty w

ch level of R

analysis of thetion

ables in parenntrols the disp

d in the “comph set of simple

he Descriptivematch the ta

t there is a sigce Difficulty fottendance.

ulty conditionsexplicate the

Same v H

=

>

will be mislea

Review

e cell means

nthesis of the play of the

pare” e effects to

es table ables

gnificant or each

s, we will needpattern of

Harder

ading

d

ObtaDiffi / emm

aining and dculty

means tables

describing

s ( pg1e2h3s b

the pairwis

by ar1y2n ) co

e simple ef

ompare ( ar1y

ffects of Re

y2n )

eview Atten

this askfor the

the ord“table” means

the varcommatest

Thein th TheRevnot

With onlpairwiseF-tests. Example Easier t

dance for e

ks for the an a2-way interac

der of the variacommand co riable specifieand tells whic

e cell means whe “Descriptiv

e F-tests tell uview AttendanEasier Practi

ly 2 Review Ae comparisons

e.

t2 = (-17.667

Ti E H S TatAm

each level o

analysis of thction ables in pare

ontrols the dis

ed in the “comch set of simp

will be the samve Statistics” a

us that the simnce is significace.

Attendance cos are redunda

/ 5.461)2 = 10

The pattern onteraction is:

Easier Practi No Review

Harder Pract No Review

Same Difficu No Review

This interactioallows us to athe main effecAttendance wmisleading

of Practice

he cell means

nthesis of thesplay of the

mpare” le effects to

me as given above.

mple effect of ant Same but

onditions, the ant with the

0.466 = F

of the

ice w > Review

tice w < Review

ulty Practicew < Review

on pattern anticipate thatct of Review

will be

e

t

t

Alte This Standtest ( So, d(smasignif An aany tANOdiffervariaterm”base

Why

1. Itpc

2. TwasL

rnative Ana

is a BG modedard Errors o(that isn’t sho

depending onaller ns) or smficant for ano

lternative is totwo cell mean

OVA error termrent design ceability, so the a” approach is

ed on the df fro

do people wh

t is based on pairwise errorcomparisons aThis approachwho attend thattend the revsignificance teLSDmmd valu

alysis of Ce

el, so all the Fof the follow-uwn in the out

whether the maller (larger n

ther!!

o use this “fulns. This is an m for BG modells, because aggregate of that since the

om all the pa

ho like this ap

the same estr term approaare significanh allows the ce review and

view and do thest, but the Cue, we would

ll Means

F-tests and fops vary with sput) that uses

cells being cons), and the s

l model error extension of els. That assthey each rethem all is thee best estimarticipants.

pproach like it

timate of variach. So, it prov

ntly different ucomparison of

do the easierhe harder praomputators wconclude the

ollow-up analysample size. s the standard

ompared havesame cell mea

term” as the the “homogensumption is thpresent a same best estima

ate is derived

t?

ability, but larvides a more sing this appf nonadjacentr practice (mectice (mean =

will give us anse two group

yses are baseWhy care? B

d error in the

e larger or sman difference

basis for comneity of varian

hat it makes smple taken froate of the variafrom using th

rger sample spowerful signroach (thought cells meansean = 44.000)= 41.667), the LSDmmd thas have equiva

ed on a singleBecause, the denominator

maller samplecan be signif

mputing an LSnce” assumpt

sense to combom different pability of each

he full design

size, and, so, nificance test,h the reverse. We might h) have the saere is no easyat we can usealent perform

e error term (M follow-up an.

e sizes, the stficant for one

SD value that tion that is mabine the withipopulations thh. The extenssample, the s

uses a smalle, and more pa can happen

have the reseme performa

y to get SPSSe to compare

mances.

MSe=111.852alyses are ba

andard error comparison

is then used ade when wen-group varia

hat all have thsion in the “fusignificance te

er standard eairwise cell mon occasion)arch hypothence as those

S to provide ththese means

2), though theased on a t-

can be largerand not

to compare compute the

ability from thehe same ll model errorest should be

error than the mean ). esis that those

who do not his s. Using the

e

r

e e

r e

e

Anotrun s

temp

SORSPLI

UNIA /DE

.

Hereanaly

her approachseparate analy

porary.

RT CASES BYT FILE LAYE

ANOVA testpeESIGN = ar1y2

e are the correysis,

h to testing simyses for each

Y pg1e2h3s. ERED BY pg1

erf BY ar1y22n.

esponding sim

mple effects th partition of t

e2h3s.

2n

mple effects F

hat shows uphe design.

F-tests from th

p in many exa

specnext an sorts splits

spec

he emmenans

amples is to u

cify that sort &nalysis comma

s the cases bys the cases b

cify DV “by”

No

Th

Th

Th

If anfufro

Fo

11

s Heefrefo

te

SOSP

U /

se the “split f

& split commaand

y the selectioby the selectio

IV (simple eff

otice that for

he SS effect (

he df-error ar

Based on csplit-analysBased on femmeans

he MSerror a

Based selethe split-anBased on temmeans

you take the nalyses and a

ull model, you om the emme

or the Easier

170.417 / 111

ere is the synffects of practeview attendaollow-ups (sinc

emporary.

ORT CASES PLIT FILE LA

NIANOVA tes/POSTHOC =/DESIGN = pg

file” option in S

ands will only

on variable on variable

fect variable)

each analysis

(ar1y2n) are t

e different

cell sample sisis approachfull design N fapproach

re different

ect portions onalysis approathe full designapproach

MSeffect fromapply the MSe

get the sameeans approac

split

1.825 = 10.46

ntax to get thetice difficulty fnce conditionce there are 3

BY ar1y2n. AYERED BY p

stperf BY pg= pg1e2h3s (g1e2h3s.

SPSS and

apply to the

s…

the same.

izes for the

for the

f the data for ach n for the

m the split error from thee F-test as ch.

66

e simple for each n, with LSD 3 groups).

pg1e2h3s.

g1e2h3s LSD)

e

Desc / emm

Alte You wthe pthe n The cthose= 7.5Hard Pleasappro

Whicand t

cribing the

means tables

rnative Ana

will sometimepattern of the n for the marg

cell means spe same N = 4555. This wouer = 66.6250

se note: Becoach taken us The emm

the effect estimated

This approgreater th

ch one to use?to be very cle

Main Effect

s ( pg1e2h3s )

alyses of Ma

es see folks omain effect. inal means is

pread the N =8 participantsuld be used to, Same = 67.5

cause this desing the emmeans analysisof review atte

d from the mooach compare non-orthog

? As you migear which one

t of Practice

) compare ( p

arginal Mea

obtain an LSDThat LSDmm

s different from

= 48 participans across just 3o compare the5000).

sign is non-omeans analysis tested and dendance and del. es the raw monality (uneq

ght expect, op you are pres

e Difficulty

pg1e2h3s )

ans of Pract

Dmmd value amd value will dm the n of the

nts across the3 conditions, e marginal m

rthogonal (has above! described thethe interactio

arginal meanual-n) of the d

pinions differ, senting.

tice Difficult

and use it to cdiffer from thee cell means.

e 6 cells, for nfor n = 16, Ueans shown t

as unequal n),

e effect of pracon. That is wh

s (without cordesign, the m

and the impo

You shoulthe same “Descriptiv67.5 for S The F-testabove, beunique comodel. Samean diffemeans of other effec

The pairwthe main e

However, interactionthose to adid not att This main because it

ty

compare the me value used t

n = 8. The maUsing n = 16 inthe “Descripti

, this analysis

ctice difficultyhy it compare

rrection for thmore these two

ortant things a

ld notice the mas the marginve Statistics” ame and 66.3

t matches whecause both antribution of taid differentlyerence amongthe groups, acts in the mod

wise comparisoeffect of Pract

Easier < H

we know fromn that this is nattended the retend the revie

effect must bt is potentially

marginal meato compare ce

ain effect of pn the LSD Coive Statistics”

s is important

y after corrected the estimat

he other effeco analyses ar

are to know w

means shownnal means froabove (50.6 f3 for Harder).

hat’s in the ANare for the corthis main effey, both are tesg the estimateafter correctindel.

ons show thetice Difficulty

Harder < Sam

m the pattern not descriptiveeview nor for

ew.

be communicay misleading.

ans, to test anell means abo

practice difficuomputator yie” table (Easier

ly different fro

ting practice dted marginal

ts in the modre likely to diff

what “your kin

n here are notom the for Easier,

NOVA table rrected or ct to the sting the ed marginal g for the

e pattern of to be:

me

of the e, neither for those who

ated carefully

nd describe ove, because

ulty spread lds LSDmmd r = 50.6250,

om the

difficulty for means –

el). The ffer!

d” expects

t

y,

Desc / em

cribing the

mmeans table

Main Effect

es ( ar1y2n )

t of Review

compare ( ar

Attendance

r1y2n )

e

As wmeamea66.5 Alsoand data Becadesiare alike auniqcont So, diffea mutest bivacrite The

Howthat Prac

Thisbeca

with the other ans shown heans from the “54 for Yes and

o, the F-test foshown below

a means show

ause there argn conditionsall collinear (nall other mult

que contributiotrolling for the

in a factorial terent than the ultiple regressa regression riate correlati

erion!

overall or ma

R

wever, we knothis is not de

ctice condition

s main effect mause it is pote

main effect, yere are not the“Descriptive Sd 55.45 for No

or “ar1y2n” inw (which matcwn in the “Des

re unequal sas, the main efnonorthogonatiple regressioon of each effe other effects

the main effe raw data masion includingweight that is

ion between a

ain effect for R

Review > N

ow from the paescriptive for tn.

must be commentially mislea

you should noe same as theStatistics” aboo).

the ANOVA ch) are not coscriptive Stati

ample sizes affects and theal, or correlateons, the modefect to the mos in the mode

cts being testarginal meansg quantitative s not the sama variable and

Review Atten

No Review

attern of the ithose in the E

municated caading.

otice that the e marginal ove (there

table above mparing the stics” above.

mong the interaction ed). Thus, el tests the odel, l.

ted are s, the same as

variables willme as the

d the

dance is:

interaction Easier

arefully,

s l