Chris Weber posted on Friday, October 29, 2010 - 3:30 pm
I am estimating a bayesian CFA and noticed that the deviance information criterion is missing when using categorical indicators? Was this intentional? I would like to compare the fit between several models and all that is included in the output is the PP p-value. Is there a particular reason why the DIC is absent? Is it not recommended for categorical indicators?
Some additional tests of fit are available in tech10. If the models are nested one possible solution is to use the simplest model (fewest number of parameters) with acceptable PPP. Nested models can also be compared using the model constraint command as on page 274 in the User's Guide.