Mark Smits posted on Tuesday, December 06, 2011 - 7:02 pm
Dear dr Muthen,
I want to compare two models containing continuous and categorical data using CFA (test estimator is WLSMV):
H1: MS by log_HA cWt log_TDL log_MP NAFL H0: MS by log_HA cWt log_TDL log_MP
However, because H1 has an additional variable, I can't test the difference in Chi2 using Difftest.
I was wondering if it is allowed to keep NAFL in the 'usevariables' in the H0-model. This would mean it is used for test statistics, but not defined in the model. I believe this would mean that NAFL is uncorrelated to the model defined in H0, please correct me if I'm wrong.
If so, I can see them as nested models and use Difftest. If not, I have to rely on qualitative comparison of the other fit parameters.