[6356] in s-news-athena
[Q] has anyone extended loess() for family=binomial?
daemon@ATHENA.MIT.EDU (Ross Gayler)
Wed Feb 1 17:40:24 1995
To: s-news@utstat.toronto.edu
From: ross@psy.uq.oz.au (Ross Gayler)
Date: 1 Feb 1995 21:11:03 GMT
The title says it all. I have a data set with 2 continuous predictors
and a binary dependent variable. When I model the outcome as a function
of 1 predictor I use gam(y~lo(x1),family=binomial). I tried to use
gam(y~lo(x1,x2),family=binomial) but ran out of RAM (I would prefer not
to reduce my sample size, but will investigate that later).
loess(y~x1*x2) works fine but the family is limited to gaussian or symmetric.
Does anyone know of an extension of loess to cope with binomial family?
I am using S-Plus for Windows V3.2
Thanks,
Ross Gayler
ross@psy.uq.oz.au
