www.cmm.bristol.ac.uk/forum

When I imputed missing values for an ordinal categorical variable (using REALCOM-impute) and then fitted the MOI to imputed data sets & combined results using Rubin's rule, all the categories (dummy variables) of that variable became insignificant. But in complete case analysis one category was significant. I need to know,

1. Does this indicate my imputations are wrong?
2. Should I fit another MOI that does not include that variable to my data set?

This is often what you might get when you do imputations - suggests that the complete cases analysis produced biased results.
Harvey Goldstein