Predictions from bayesian logistic model

[Geoff Dougherty]

Hi, all.

I am working with a dataset that describes diabetes status (0 or 1) for individual residents of multiple states, across multiple years. I'm interested in obtaining age/sex/race adjusted estimates of prevalence for state/year combinations that have some shrinkage to account for small sample size, so I've specified a crossed effects model:

melogit diabetes age sex race || _all: R.year || _state:

predict re*, reffects

While the model ran successfully, I was not able to obtain predictions in a reasonable amount of run time. I subsequently came across some info suggesting fully Bayesian crossed effects models are more tractable, so I specified the following:

 bayes, saving(output/bayes, replace): melogit diabetes age sex race || _all: R.year || _state:

This model has run, but I am not really understanding how to obtain estimates of the random effects for state and year. They do not seem to be in the output from -bayesstats- or in the output file.

Thanks for any suggestions!

[Geoff Dougherty]

Update: It appears I can access the estimates I need via -bayes, showreffects-. However, I need to be able to either update the existing data file with those estimates, or output to a separate .dta. -estout- does not seem to output the right estimates in this instance.