t-test for subset of one variable against second variable

[Kate Doyle Feingold]

I have a randomized control trial with treatment and control taking survey 1, then treatment watches video and takes survey 2.

I have a variable called group that contains two subsets - treatment and control. I have two other variables, survey_1 and survey_2. I need to run a ttest comparing the mean of the control group survey_1 to the mean of survey_2.

[Joseph Coveney]

You'll have to make a boatload of assumptions about confounding, but you can try a so-called cell means model. Usually it's used in a factorial ANOVA context, but in your case, it's in a repeated-measures context, and so you'd use it in a mixed-effects model. Mabye something like the following.

.ÿ
.ÿversionÿ15.1

.ÿ
.ÿclearÿ*

.ÿ
.ÿsetÿseedÿ`=strreverse("1523828")'

.ÿ
.ÿquietlyÿsetÿobsÿ14

.ÿgenerateÿbyteÿpidÿ=ÿ_n

.ÿgenerateÿdoubleÿpid_uÿ=ÿrnormal()

.ÿ
.ÿgenerateÿbyteÿtrtÿ=ÿmod(_n,ÿ2)

.ÿlabelÿdefineÿGroupsÿ0ÿControlÿ1ÿExperimental

.ÿlabelÿvaluesÿtrtÿGroups

.ÿ
.ÿquietlyÿexpandÿ2

.ÿbysortÿpid:ÿgenerateÿbyteÿsrvÿ=ÿ_n

.ÿlabelÿdefineÿSurveysÿ1ÿ"Surveyÿ1"ÿ2ÿ"Surveyÿ2"

.ÿlabelÿvaluesÿsrvÿSurveys

.ÿ
.ÿgenerateÿdoubleÿrspÿ=ÿpid_uÿ+ÿrnormal()

.ÿquietlyÿreplaceÿrspÿ=ÿ.ÿifÿtrtÿ==ÿ"Control":Groupsÿ&ÿsrvÿ==ÿ"Surveyÿ2":Surveys

.ÿ
.ÿ//ÿ"Cellÿmeansÿmodel"
.ÿxtregÿrspÿ0.trt#1.srvÿ1.trt#1.srvÿ1.trt#2.srv,ÿi(pid)ÿre

Random-effectsÿGLSÿregressionÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿNumberÿofÿobsÿÿÿÿÿ=ÿÿÿÿÿÿÿÿÿ21
Groupÿvariable:ÿpidÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿNumberÿofÿgroupsÿÿ=ÿÿÿÿÿÿÿÿÿ14

R-sq:ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿObsÿperÿgroup:
ÿÿÿÿÿwithinÿÿ=ÿ0.0003ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿminÿ=ÿÿÿÿÿÿÿÿÿÿ1
ÿÿÿÿÿbetweenÿ=ÿ0.0219ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿavgÿ=ÿÿÿÿÿÿÿÿ1.5
ÿÿÿÿÿoverallÿ=ÿ0.0203ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿmaxÿ=ÿÿÿÿÿÿÿÿÿÿ2

ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿWaldÿchi2(2)ÿÿÿÿÿÿ=ÿÿÿÿÿÿÿ0.28
corr(u_i,ÿX)ÿÿÿ=ÿ0ÿ(assumed)ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿProbÿ>ÿchi2ÿÿÿÿÿÿÿ=ÿÿÿÿÿ0.8711

----------------------------------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿrspÿ|ÿÿÿÿÿÿCoef.ÿÿÿStd.ÿErr.ÿÿÿÿÿÿzÿÿÿÿP>|z|ÿÿÿÿÿ[95%ÿConf.ÿInterval]
-----------------------+----------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿtrt#srvÿ|
Experimental#Surveyÿ1ÿÿ|ÿÿÿÿ.445805ÿÿÿ.9058157ÿÿÿÿÿ0.49ÿÿÿ0.623ÿÿÿÿ-1.329561ÿÿÿÿ2.221171
Experimental#Surveyÿ2ÿÿ|ÿÿÿ.4656917ÿÿÿ.9058157ÿÿÿÿÿ0.51ÿÿÿ0.607ÿÿÿÿ-1.309675ÿÿÿÿ2.241058
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ_consÿ|ÿÿ-.4708248ÿÿÿ.6405085ÿÿÿÿ-0.74ÿÿÿ0.462ÿÿÿÿ-1.726198ÿÿÿÿ.7845487
-----------------------+----------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿsigma_uÿ|ÿÿ1.4213813
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿsigma_eÿ|ÿÿ.95469132
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿrhoÿ|ÿÿ.68911664ÿÿÿ(fractionÿofÿvarianceÿdueÿtoÿu_i)
----------------------------------------------------------------------------------------

.ÿ
.ÿtestÿ0.trt#1.srvÿ=ÿ1.trt#2.srv

ÿ(ÿ1)ÿÿ0b.trt#1b.srvÿ-ÿ1.trt#2.srvÿ=ÿ0

ÿÿÿÿÿÿÿÿÿÿÿchi2(ÿÿ1)ÿ=ÿÿÿÿ0.26
ÿÿÿÿÿÿÿÿÿProbÿ>ÿchi2ÿ=ÿÿÿÿ0.6072

.ÿ
.ÿexit

endÿofÿdo-file


.

[Kate Doyle Feingold]

ok, thank you.