Question Earthority!

“Data imputation can be a good idea. Sometimes data can be missing and it can be replaced without screwing up the analysis. For example, if we are trying to understand travel behavior and income is missing, and if it can be predicted in a half-way decent way, then the researcher doesn’t have to toss the observation. But it almost always a bad idea when the missing variable is the object of study. For example, if income is missing and we’re trying to predict income there is a pretty good chance that the predicted income value will be way off the mark.” (07/31/06)