Tuesday, October 06, 2009

Best Methods Question Ever

I read PSJR once a week, and it rarely fails to amuse. This week, there was a gem of a question hiding in the rocks:

I'm doing a logit estimation in Stata and want to generate predicted values afterward. The problem I'm finding is that my dataset includes cases that are blank values for the dependent variable -- and when I enter the "predict" command, Stata imputes predicted values for those blank cases. I don't want this, but can't figure out a way to tell Stata to just generate predicted values for cases in which there's a value for the dependent variable. Any thoughts?

A helpful person was able to think of an answer, after long (I assume) deliberation:

Delete the cases with missing values on the dependent variable before you estimate the model.

This is, I submit, sound advice. Of course, another later commenter also has good advice:

Become a qually. It's not too late yet.

Again, on the mark. For quallies, confused obscurantism passes for profundity.

No comments: