Page 1 of 1

Measurement Invariance testing with CCMM models

Posted: Tue Apr 02, 2019 2:15 pm
by PiaKreijkes
Hi,

Is it possible to test measurement invariance using a CCMM model? I have 5 constructs measured by 30 items that are each assessed at two time points. I want to create factor scores and compare Time 1 and Time 2 scores in a later model so need to establish measurement invariance first. I wanted to do this in Mplus but cannot model my structure adequately there.

I have time (level 1) in students (level 2) in classes (level3) and teachers (level 4). Some students changed classes over time and some classes are taught by multiple teachers and some teachers teach multiple classes.

Can I model this in MLwiN and if so, are there any resources that discuss this?

Thanks,
Pia

Re: Measurement Invariance testing with CCMM models

Posted: Thu Apr 04, 2019 7:16 am
by billb
Hi Pia,
MLwiN has some functionality for multilevel factor analysis (using MCMC) and I suspect it could do cross-classified/multiple membership as well though I have never tested it so I wouldn't vouch for it. I think if you construct factor scores elsewhere and want to fit a more standard multivariate response CCMM model to those scores (i.e. a 2 step process of create factor scores assume they are true and treat them as data in the CCMM model) then that should be possible though I don't think we have any resources there describing such models - probably best to look at the literature on such models and see what software has been used.
Best wishes,
Bill.

Re: Measurement Invariance testing with CCMM models

Posted: Fri Apr 05, 2019 10:46 am
by PiaKreijkes
Hi Bill,

Thanks for your reply. I usually do my multilevel factor analysis in Mplus and can also do invariance testing there but unfortunately it cannot take into account that the group membership changed between the two timepoints for many students. I could not find any papers who did this either. Maybe choosing the more dominant group as the cluster might be a compromise in that case.

Just a related questions. Does MLWin always require the long format when different time points are involved?

Thanks,

Pia

Re: Measurement Invariance testing with CCMM models

Posted: Fri Apr 05, 2019 5:18 pm
by billb
Hi Pia,
It depends on what model you are fitting. If you look at the MLwiN manuals for MV models you will see that the data is in wide format and then when you specific the columns for the different responses the software constructs the long format in the background.
Best wishes,
Bill.

Re: Measurement Invariance testing with CCMM models

Posted: Mon Apr 08, 2019 8:11 am
by PiaKreijkes
Thank you!