I have achievement data for about 600 students. About 60% of the data are missing by design due to multi-matrix sampling. Now, is there a way to perform CFA (invariance tests) in Mplus using my whole sample despite that big amount of missing data? And what would the way to accomplish that kind of analysis?
Any suggestions would be of great help to me. I thank you in advance.
Although nobody takes all items, if all students take the same reduced number of items, you can do a multiple-group analysis, by which it is then possible to do invariance testing. If you are working with categorical items then multi-group analysis is accomplished via the KNOWNCLASS option.
I am running a mixture model using PISA data which has large numbers of data missing by design (matrix sampling). Do you know what is the effect of matrix sampling on the estimation of latent classes? Also, what is the effect on the identification of the latent classes when everyone did not take the same items (examinees take different booklets).