(This article was originally published at Carlisle Rainey » Methods/Statistics, and syndicated at StatsBlogs.)

I've discussed imputing missing outcome variables before, but I've realized a new problem for those who oppose it. If you are opposed to multiply imputing an outcome variable, then you have to be opposed to duration models that adjust for time censoring. You have to oppose models such as the tobit that deal with truncated observations. You also have to oppose partial observability models. Each of these models is dealing with a specific form of missingness in the outcome variable. The goal is exactly the same as multiple imputation and the mathematics is very similar. You cannot oppose multiple imputation of outcome variables and then use these models. Simply list-wise delete and move ahead.

**Please comment on the article here:** **Carlisle Rainey » Methods/Statistics**