Dear listers, I am trying to use an old script which was working well in the previous R version. It looks like if it no longer works in R.6.0. I have a model of the form: glm(nath2$Positif ~ n + yearday + x + y + I(x^2) + I(y^2) + yearday:x + yearday:y, family = poisson, data = nath2) and want to get predicts from a data.frame whose column names are: > names(data1) [1] "x" "y" "yearday" "n" x and y are geographical coordinates, yearday is equal to 120 and n equal 100; they are all numerics: > sapply(data1,is.numeric) x y yearday n TRUE TRUE TRUE TRUE when I use the function predict: > z1<-predict(mod1b,newdata=data1,type="response") Error: variables ?I(x^2)?, ?I(y^2)? were specified with different types from the fit Any idea about what goes wrong ? Patrick