Posted also to sci.stat.edu, where the same question appeared. On Thu, 16 Dec 1999 00:03:58 +0100, EikeRietzel <[EMAIL PROTECTED]> wrote: > Hi, > I have a very urgent problem concerning GLM (repeated measures). > > The Mauchly sphericity test prints out: > Mauchly-W = 0.000 > approx-chi2 = . > df = 9 > significance = . > > Does this mean the within-subject factors fail to meet the assumption of > sphericity? What is clear is that there is something unusual that made W come out 0.0. My guess would be that somewhere you have variances or covariances that are *exactly* equal or exactly 0 owing (probably) to discrete data; under the assumption of continuous data, there would never be a W of 0.0. So, what is odd about your data? The problem of having no chi2 or no "significance" for the missing chi2 follows from that. By the way, the test (if I remember right) is a pretty fuzzy indicator --not to be relied upon in either direction. If you are going to be concerned with those issues, you need to look more closely than that. http://www.pitt.edu/~wpilib/index.html