Dear All, I have two treatment levels and multinomial behaviors as response and also a covariate (size). The individuals were split to the two treatment levels on the first day and switched on the second day. I am interested not only the treatmen effect but the effect of the day, such that the individuals experinced treatment 1 first, or treatment 2 first. I conditioned on the most frequent behavior and than run separate poisson models on the rest of the behaviors (proportions of time when they did a behavior given that they do not do the most frequent). glm(behav1~treatment*day+size, family=poisson(), data=data) I wonder that it is not correct because the day effect is nested in the treatment effect but I do not know how set it up in glm(). Also do I need to do any corrections because of the separate tests on each beahvior? It is more a statistical question. Thanks for any help in advance! Katalin