The survival package is one of the few with a default of model=FALSE.?? There are a couple of reasons for this, one major one is that I work in medical research and making unnoticed copies of the data is a risk to patient confidentiality, e.g.,? multi-state models all have an "id=" argument.?? (The primary one, however, is probably that I'm stubborn.) In any case, this would argue for making weighted.residuals an S3 method.? The survival NAMESPACE file currently has 200 S3 methods, so what's one more? PS. An addition I would like to see in the Writing R Extensions manual would be a discussion of how to find a list of all the S3 methods which one's current package might need, or be useful to add.?? A good fraction of those in the survival package were in response to user email of "why don't you have a method for? XYZ", when I was completely unaware of the existence of S3 method XYZ. At one time I did a search for UseMethod in the R source tree, but that doesn't find things like xtrfm(), which is defined using .Primitive. Terry T.
Terry M Therneau, PhD Department of Quantitative Health Sciences Mayo Clinic therneau at mayo.edu "TERR-ree THUR-noh" [[alternative HTML version deleted]]