topic Re: Is it possible to generate a pooled -2 Log Likelihood from proc mianalyze? in Statistical Procedures

Is it possible to generate a pooled -2 Log Likelihood from proc mianalyze?

ejamro — Wed, 17 Nov 2021 21:36:15 GMT

I am using proc mi to impute missing values, proc glimmix for regression, then proc mianalyze to pool the estimates from the imputed datasets. I receive the imputation-level -2 Log L with the glimmix results, but is it possible to obtain a pooled -2 Log L from mianalyze?

Re: Is it possible to generate a pooled -2 Log Likelihood from proc mianalyze?

SAS_Rob — Wed, 17 Nov 2021 21:56:15 GMT

Because the theory behind the combining of estimates from multiply imputed data requires both a point estimate and a standard error (which you don't have in this case), you cannot get a pooled -2 LL measure using MIANALYZE. I have not seen a discussion anywhere in the literature as to if or how this might otherwise be done, but maybe someone else in the Community has some suggestions.

Re: Is it possible to generate a pooled -2 Log Likelihood from proc mianalyze?

SteveDenham — Tue, 23 Nov 2021 18:14:20 GMT

A quick google on "combining log likelihoods" gives a lot of interesting results ranging from as simple as adding the log likelihoods (equivalent of multiplying the likelihood) which assumes independence between the imputations, to several Bayesian methods, and on to some good Q and A on stackexchange.

The one thing I thought was a common thread is that there is good theory behind the adding of log likelihoods/multiplying likelihoods when you think of it as combining probabilities. There is a strong independence assumption though.

SteveDenham