Mahdi posted on Thursday, April 24, 2014 - 7:44 pm
Dears Dr.Muthen Hi What is your suggestion for running EFA (CFA or perhaps SEM)for non-independent observations data with many clusters (about 300) with 900 observations as family data? We have 12 continuous indicators and my purpose is considering the family members correlations in modeling. Please introduce reference papers. Thank you very much for your good help. Mahdi
You can take into account non-independence of observations due to clustering using TYPE=COMPLEX or TYPE= TWOLEVEL, THREELEVEL, or CROSSCLASSIED. The introduction the Chapter 9 of the user's guide discusses these approaches. If you don't have a lot of family members, you can consider a multivariate approach which is discussed in the Topics 7 and 8 course handouts an videos on the website.