Rongjin Guan posted on Wednesday, October 21, 2015 - 11:49 am
I have two general questions regarding running LCA:
1. I have a warning message and wonder if this needs to be addressed:
*** WARNING in MODEL command All variables are uncorrelated with all other variables within class. Check that this is what is intended.
2. I have a survey data, with >120 items (questions). Answers are 1-6 likert style. I used EFA and CFA to reduce the number of items to 48, which belong to 9 subscales. Now I want to run LCA with these 9 subscales, shall I use the sum of the items or the mean of the items within the subscale as input to LCA?
I tried with Sum of items and also mean of items, and the results are different. So I am wondering which one is the correct practice.
2) This should not matter, but using a mean keeps the variance lower which can be helpful. You should check your analysis carefully to see why you get a difference, e.g. check that you have the same sample size and that the best logL is replicated.