Sapsi wrote: > > Hi, > I have sample of 40,000 observation grouped into 2 response groups > (responders = 1000). Number of explanatory variables=55 > While using prog logistic, i get possible quasi complete separation. > When i drop to a subset of 50 this problem goes, again a different set > of 50 brings back this problem! > 0) Why does this problem happen? > 1) Is there a way to detect which variables cause this problem? > 2) How can i get rid of it? > > Any help and detailed help will be **really** appreciated.
It's easiest to describe in terms of a single predictor. Complete separation means Y=0 for X<k and Y=1 for X>k for some k (or vice-versa, that is the 0 and 1 can be swapped). The fitting process breaks down as the program tries to fit a step function. Quasi CS is the same *except* Y can be 0 or 1 at X=k. . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
