- Joined
- Jun 21, 2003
I'll wait and see what Mathman will say. We may ask him to give up his handle should he give us a wrong answer.
*Ahem* Strange to say, my last PhD student wrote his dissertation on strategies for dealing with missing data in the setting of investigations in the social sciences. Statistically, the problem with just filling in all the blanks with the same dummy number (the mean of the actual responses, for instance), is that this artificially lowers all the variances, making the study appear to be more robust than it really is. Results will satisfy standard criteria for statistical significance, when in reality they are baloney.
(This consideration is swamped, however, by the fact than people simply lie on questionnaires. )