I have been using proc glimmix for longitudinal analysis with dichotomous outcomes (4 time points, separate models for different outcomes). My odds ratios appear to be overestimated when time is included as a covariate in the model. Could this be a product of having a very time dependent outcome?
How are you incorporating time? If you are doing GEE type models (R side), you will bias estimates away from the extremes. Try fitting a G side model, with method=laplace or method=quad.
See Stroup's Generalized Linear Mixed Models for examples.
Steve Denham
Available on demand!
Missed SAS Innovate Las Vegas? Watch all the action for free! View the keynotes, general sessions and 22 breakouts on demand.