s-news
[Top] [All Lists]

offset and weight term

To: <s-news@lists.biostat.wustl.edu>
Subject: offset and weight term
From: "D'Angelo, Gina" <dangelo@upci.pitt.edu>
Date: Wed, 26 May 2004 11:17:57 -0400
Thread-index: AcRDNVnJgJif4lDKSDKIGQGH3CEWWA==
Thread-topic: offset and weight term
Dear S-plus users,

I have an inquiry about the weight and offset options in glm and coxph.  I 
attempted to locate details of estimator calculations from S-plus but was 
unsuccessful.  Does anyone know how S-plus is estimating the parameter and 
variance estimates when including offset terms and weights.  I am solving a 
missing data problem by incorporating either weights or offset terms.  I know 
there have been a number of variance calculations proposed so I am concerned if 
S-plus is using the one I'm interested in.  For weighting does S-plus just 
weight the score and then estimate the parameters.  Is the robust variance the 
variance estimate.  If anyone knows which references were used to write the 
code that would be very helpful.

I am using the following commands:
glm(newdat.1[,1]~newdat.1[,2], family=binomial,weights=(1/p.2))
coxph(Surv(y3,delta3)~x3+v3,weights=weight.p.use.mar,method="breslow")

glm(newdat.1[,1]~newdat.1[,2]+offset(newoffset), family=binomial)
coxph(Surv(y3,delta3)~x3+v3 + offset(offset),method="breslow")

Thanks for any input you have.

Gina


<Prev in Thread] Current Thread [Next in Thread>