Re: [S] Potentially serious bug in S-PLUS 4.5

David O. Nelson (
Wed, 17 Jun 1998 12:03:19 -0700

Indeed it would have been nice (not to mention professional on the part
of Mathsoft) to be warned of this, as I spent a day beating my head
the wall trying to port some WORKING S code from Unix to S-PLUS 4.5,
chasing discrepancies between predict.glm() in the two systems. As it
my first experience with S-PLUS for windows, the conclusion that I drew
to be very careful about recommending S-PLUS to anyone but the most
of programmers. I thank Prof. Ripley for alerting us, and am thinking
wistfully) of sending a bill to Mathsoft for my wasted time.

Prof Brian D Ripley wrote:

> The problem is a new argument drop.unused.levels in model.frame.default, which
> defaults to T. (Despite the help page, it is the fourth argument, not the
> third). The explanation says that without this there may be singular model
> matrices. However, forming the model matrix is the job of model.matrix, not
> model.frame, and singularity is only important when fitting, not predicting.
> I have no idea how far-reaching this is: to find out one would have to examine
> all uses of model.frame throughout the system _and_ all user-contributed code.
> I do know that I _usually_ get wrong answers with predict.tree, as this maps
> the factors to their codes. I suspect that this applies to most
> model-fitting functions that use model.frame.
This message was distributed by To unsubscribe
send e-mail to with the BODY of the
message: unsubscribe s-news