Step Summary

Multinomial Logistic Regression supports pooling of regression coefficients; however, you will note all tables in the output show the results for each imputation and the original data. This is because the file is split on Imputation_, so all tables that honor the split variable will present the split file groups together in a single table.

You will also see that the Parameter Estimates table does not show pooled estimates; to answer why, look at the Step Summary. We requested stepwise selection of model effects, and the same set of effects was not chosen for all imputations, thus it is not possible to perform pooling. However, this still provides useful information because we see that ed (Level of education), employ (Years with current employer), marital (Marital status), and address (Years at current address) are frequently chosen by stepwise selection among the imputations. We will fit another model using just these predictors.

Next