[Home Page] [Previous Page]
[Old Roundtable Archives] [New Roundtable]
Posted by John Taylor on January 27, 2000 at 12:21:29:
In Reply to: weighting posted by Val Keel on January 27, 2000 at 06:25:59:
Yes. The 250 being weighted "up" would receive wieghts of about 1.94 while those weighted "down" would have weights of about .06. The risk in weighting is that some individuals who, happen to have unusual answers, receive large weights. My own rule of thumb is be very cautious with weights above 2.5 and never go above 3.5.
Of course, this recommendation assumes the maldistribution (is there such a word?) in your sample was intentional (e.g., you oversampled customers so you could read their results separately) AND the over- and undersampled groups have different answers on important questions.
Subject: