I am using glmmPQL on a dataset which has no missing values: antgl<-glmmPQL(SF~Proc*Min, random= ~1|Trial, family="binomial") I get the error message: iteration 1 Error in na.fail.default(list(invwt = c(1.36892171002990, 1.36680104392581, : missing values in object I also get this error if na.action is set na.fail. If I set na.action to na.omit to ignore the (mythical) missing values, I get other error messages. There are no missing values - I have used is.na(antgl) to check this. Can anyone explain why it seems to think there are? Thanks