empty model in mulitple imputations framework (mixed models)

[olivier dejardin]

Dear all,

I am currently analyzed data with missing data.

In complete case analysis framework, the xtmelogit command provides a crucial information on the last line in the random effects table (the "LR test VS. logistic model" statistic). Unfortunately, using the mi estimate command before the xtmelogit command, this information is no more provided.

I have try to perform some LR tests using lrtest command and milrtest command ... but with no result

Does anyone know how I can test the empty model with multiple imputation?

Many thanks,
Olivier

[Federico Tedeschi]

I am facing the same problem. Judging from what I read here: https://www.stata.com/meeting/fnasug...eiros_2008.pdf , at page 4, the problem is how to combine LR tests from several models. You can get the results of each imputed dataset by using the "saving" option with "mi estimate", and then the "estimates use" command, with the "number" option. In this way, you can see each of the M estimates in single datasets. To get the global estimates, however, you have to use number "M+1", and in that case you'll find again no LR test.