I am running a path analysis with two exogeneous continuous variables, one continuous mediating variable and a dichotomous outcome variable. Furthermore, I am doing a multiple group analysis because I want to compare men and women. In order to be clear, I use weights (I saw that it is not possible to use all estimators when weighting your data).
Which estimator do I have to use? At this moment, I use WLSMV. I read the manual and I decided that this was the best method to use. With a colleague, however, I had a discussion whether ML would be better or not. Wheras WLSMV creates probit estimates, ML creates logit estimates as I understood. Now I am confused about which is the best estimator to use in this particalur situation?
Weights can be used with both weighted least squares estimation and maximum likelihood estimation. You cannot use the ML estimator option with weights because with weights the standard errors are not maximum likelihood. You can use MLR which gives the same parameter estimates as ML. So you choice is whether you want probit or logistic regression.