[19269] in s-news-athena
[S] GAM interaction terms
daemon@ATHENA.MIT.EDU (Cande V. Ananth)
Mon Sep 20 11:27:22 1999
Date: Mon, 20 Sep 1999 11:17:49 -0400 (EDT)
From: "Cande V. Ananth" <ananthcv@epi.umdnj.edu>
Message-Id: <199909201517.LAA14803@epi.umdnj.edu>
To: s-news@wubios.wustl.edu
I'm attempting to fit a Poisson regression model to with 2
continuous variables, X1 and X2, using the gam() function.
I'm allowing smooth terms for X1 and X2. The data contain
about 80,000 observations. The code that I have is:
gam.fit1 <- gam(y ~ s(X1, degrees=2, span=1/2) +
s(X2, degrees=2, span=1.2),
family=poisson, na.action=na.omit)
1. I need to allow an interaction term for s(X1) and s(X2).
How can I do that in gam() ? Moreover, even simple models
in gam(), [gam(y ~ s(X1), family=poisson)] given the size
of my data, takes about 30 minutes to converge !
2. Are there any papers/books/unpublished materials/etc that
explains the gam() functions, or even a "case-study" which
provides the gam() codes used for data analysis ? I have
looked at the SPlus manuals, and the Chambers & Hastie
books.
I'm using SPlus 3.4 on the UNIX with the Solaris OS. My
machine has 512 megs memory, and runs on a 333 mHz processor.
I'd appreciate any help. Thanks.
--
Cande Ananth
ananthcv@EPI.UMDNJ.EDU
UMDNJ, Dept of OB/GYN
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news
