Hi, For efficiency purposes I'm using glm.fit instead of glm, and I'm pre-constructing a model design matrix using model.matrix(). I'm finding that the 'NAs' in the x are being removed by model.matrix by default, and would like to know how could I keep them (or remove the corresponding NAs in the Y) so that the length for both response and predictor would be the same. Thanks for your reply. myformula = ~ offset(Bgrid[i,1] * x) MM <- model.matrix(myformula) alpha[i] <- coef(glm.fit(MM, Y, family=binomial())) Ted ------------------------------------- Ted Chiang Bioinformatics Analyst Centre for Computational Biology Hospital for Sick Children, Toronto 416.813.7028 tchiang at sickkids.ca