Model fit statistics after using mi estimate

[Al Bothwell]

I am a bit of a quandary. I used multiple imputation and then run this command:

mi estimate: logit outvar predvar

Now I want model fit indices (AIC, BIC) but can't obtain them with imputed data. Any suggestions?

[Rich Goldstein]

AIC and BIC are for maximum likelihood estimates; mi estimate is not a maximum likelihood estimator; that said, you could estimate the model for each imputed data set, obtain AIC (etc.) for that estimate and then average them over the entire set - not sure it would be meaningful, however

[Al Bothwell]

Thank you. I run separate logistic models with and without imputation. The results are very similar, so I went along with the complete case dataset.