Carla Bann posted on Tuesday, June 26, 2007 - 3:53 pm
I am running a latent profile analysis. I would like to compare a 4 class model assuming equal variances for the indicator variables to one allowing the variances to be freely estimated. Is there a way to determine whether freeing the variances significantly improves the model fit? For example, would it be appropriate to compute a chi-square difference test using the values for the 2 models listed as "Loglikelihood H0 value?" Thank you!