Alex Chavez posted on Thursday, August 19, 2010 - 11:35 am
I have 3 nominal responses (each with three levels) from different experimental conditions and J covariates for each of 32 subjects (32 x 3+J matrix). I have previously modeled this data in R using a multinomial logistic model with a random subject-level intercept. My goal is to make inferences about the probabilities of each response level across conditions and to test for an overall effect of condition, while controlling for within-participant correlation. Is using the CLUSTER option in Mplus on participant IDs a viable approach for the model, and would this estimate random intercept variances? Are there other models I might consider?