SEM - goodness of fit indices

[Elizabeth Mumford]

Greetings -
Whether I write my syntax directly or use the model building tool, I am not getting the goodness of fit options I would expect. First, there is no LR test of model vs. saturated (chi2 test) output at the end of the model.

Second, I somehow seem to be running the model with robust errors, because I am unable to produce more than SRMR and the CD following the model using
estat gof , stats(all)

If I specify other options, this is the result:

estat gof, stats(rms)
Note: model was fit with vce(robust); only stats(residuals) valid.

Per a 2014 posting and response (http://www.statalist.org/forums/foru...ta-fit-indices), the problem may be that I am using survey weights with my data, and can only expect to get the SRMR and the CD?
Thank you --

[Richard Williams]

A lot of stuff that works with non-svy is not appropriate for svy: analysis. The following isn't sem-specific but it does go over some of the issues.

http://www3.nd.edu/~rwilliam/xsoc73994/SvyCautionsX.pdf

If you want to use GOF measures to assess ways in which your model could be improved, consider running without svy. Or, check out some of the other estat commands, like estat mindices and estat scoretests.

[Elizabeth Mumford]

Thank you kindly. In your linked guidance (http://www3.nd.edu/~rwilliam/xsoc73994/SvyCautionsX.pdf ) I see that the AIC and BIC tests would require an underlying maximum likelihood assumption that inclusion of survey weights does not allow. I excluded the pweights to see if that was truly the block, and it was. The RMSEA and CIF results however indicate a poor model fit (whereas the SRMR was acceptable), and I am not sure if that is a result of the weak model, or the exclusion of the weights.