q about interactions
Posted: Fri Jul 26, 2013 6:42 pm
I have an interaction between an unordered categorical variable and a continuous variable which has some missing values. If I am using the JAV approach to impute the missing values for the interaction, I understand that I need to manually create the variable first in Stata and then use it as a response variable in realcomImpute. In terms of manually creating the variable, is this just a matter of multiplying the unordered categorical variable by the continuous variable? I understand that if it is a response variable, you can't include separate dummy variables in realComImpute, such as continuous*category_1 continuous*category 2 etc.