Nlm

Hi

I use both R and Splus to fit miximum likelihood models.  In general these
are Gaussian mixture models of various kinds.

The ms (in Splus) and nlm (in R) generally provide us with the same results
However nlm is not nearly as stable and reliable with respect to finding the
min of a surface.  Nlm on the other hand is faster.  We have played with the
various tol controls but have been unable to increase reliability.  Do you
Have any suggestions?

Thanks

Tim Gage
Professor Anthropology and Epidemiology
University at Albany
Albany NY.

tbg97 at albany.edu