1. Is there ever a situation using the R recommended packages only (to keep things reasonable) where one explicitly needs to create indicator variables to do statistical modeling? 2. If so, does model.matrix() always suffice to do this? In any situation I can imagine, one would create a factor and then proceed from there. But I may lack imagination. Thanks. -- Bert -- Bert Gunter Genentech Nonclinical Biostatistics Internal Contact Info: Phone: 467-7374 Website: http://pharmadevelopment.roche.com/index/pdb/pdb-functional-groups/pdb-biostatistics/pdb-ncb-home.htm