Hello everyone, I want to fit a following model as an alternative to log binomial model (Thomas Lumley,et al?2006). But I come across with this error: ? Error in eval(expr, envir, enclos) : cannot find valid starting values: please specify some? geeglm (outcome ~ treatment, data = dat, family=gaussian(link="log"), id=subject_id) Using SAS (Genmod) I can run this model without any problem: proc genmod data=dat DESCENDING; model outcome = treatment / dist=normal ?link=log; repeated subject=subject_id; run; Could anyone offer some suggestions? Thank you Leila [[alternative HTML version deleted]]