I am running a CFA in Mplus with 12 categorical (4 response categories) indicators to produce 2 factors. Should I be testing the indicator variables for multivariate normality? And if so, is there a test in Mplus for this?
I am trying to perform a measurement invariance test for a latent factor being loaded by 4 categorical ordinal variables. Each of the categories has 11 groups and MPLUS7 imposes continuous variable methodology. But the latent class is not multivariate normal under the MLR estimator. Plus, there is no difference in chi-square test statistic produced by the ML and MLR estimators. How do I address the lack of multivariate normality in the latent factor?