On Nov 9, 2009, at 11:15 PM, Jens Levy wrote:
I am interested in evaluating whether or not the joint effect of two
independent variables on an outcome are additive. I want to be able
to give
something analogous to the Breslow-Day test of homogeneity, which
evaluates
effect modification on the multiplicative scale, but on the additive
scale.
I would also like to calculate the common Risk Difference across two
strata.
Does anyone know of any R package that has these functions, which
focus on
additive relationships in stratified analyses?
It would be possible to use glm with a quasi() family argument with
link="identity" and variance="mu". This would make the effect
relationships linear in their untransformed scales but allow the
variance to be Poisson.
The mgcv package is the one suggested by Simon Wood for what might be
called "basic" generalized additive models described in his book in
chapter 5, but there are a large number of other packages that handle
generalized additive models of various sorts.
http://search.r-project.org/cgi-bin/namazu.cgi?query=%22additive+models%22&max=100&result=normal&sort=score&idxname=functions&idxname=Rhelp08&idxname=views
--
David Winsemius, MD
Heritage Laboratories
West Hartford, CT
_______________________________________________
R-sig-Epi@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-epi
