R help - Jun 2004 - poisson regression with robust error variance ('eyestudy')

Dear all,

i am trying to redo the 'eyestudy' analysis presented on the site
http://www.ats.ucla.edu/stat/stata/faq/relative_risk.htm
with R (1.9.0), with special interest in the section on "relative risk 
estimation by poisson regression with robust error variance".

so i guess rlm is the function to use. but what is its equivalent to the 
glm's argument "family" to indicate 'poisson'? or am i
somehow totally
wrong and this is not applicable here?

thx a lot-
lutz


============================Lutz Ph. Breitling, CMd
Unit? des Recherches M?dicale
H?pital Albert Schweitzer
B.P. 118 Lambar?n? (GABON)  
-------------- next part --------------

---

On Wed, 2 Jun 2004, Lutz Ph. Breitling wrote:
> Dear all,
>
> i am trying to redo the 'eyestudy' analysis presented on the site
> http://www.ats.ucla.edu/stat/stata/faq/relative_risk.htm
> with R (1.9.0), with special interest in the section on "relative risk
> estimation by poisson regression with robust error variance".
>
> so i guess rlm is the function to use. but what is its equivalent to the
> glm's argument "family" to indicate 'poisson'? or am
i somehow totally
> wrong and this is not applicable here?
>
No, no.  You want glm() and then a function to compute the robust
covariance matrix (there's robcov() in the Hmisc package), or use gee()
from the "gee" package or geese() from "geepack" with
independence working
correlation.

These are not outlier-resistant estimates of the regression coefficients,
they are model-agnostic estimates of the standard errors.

Stata is unusual in providing these covariance matrix estimates for just
about every regression estimator.  I think R should consider doing
something similar, but haven't got around to it.

	-thomas

> thx a lot-
> lutz
>
>
> ============================> Lutz Ph. Breitling, CMd
> Unit侀 des Recherches M侀dicale
> H侓pital Albert Schweitzer
> B.P. 118 Lambar侀n侀 (GABON)
>
Thomas Lumley			Assoc. Professor, Biostatistics
tlumley at u.washington.edu	University of Washington, Seattle

The below is an old thread:

On 02-Jun-04 10:52:29, Lutz Ph. Breitling wrote:
> Dear all,
> 
> i am trying to redo the 'eyestudy' analysis presented on the site
> http://www.ats.ucla.edu/stat/stata/faq/relative_risk.htm
> with R (1.9.0), with special interest in the section on "relative
> risk estimation by poisson regression with robust error variance".
> 
> so i guess rlm is the function to use. but what is its equivalent
> to the glm's argument "family" to indicate 'poisson'?
or am i
> somehow totally wrong and this is not applicable here?
> 
> thx a lot-
> lutz
> ============================> Lutz Ph. Breitling, CMd
> Unit? des Recherches M?dicale
> H?pital Albert Schweitzer
> B.P. 118 Lambar?n? (GABON)
It seems it may have led to a solution. However, I have tried
to trace through the thread in the R-help archives, and have
failed to find anything which lays out how a solution can be
formulated.[*]

I'm interested in the same question. Basically, if I fit a GLM
to Y=0/1 response data, to obtain relative risks, as in

  GLM <- glm(Y ~ A + B + X + Z, family=poisson(link=log))

I can get the estimated RRs from

  RRs <- exp(summary(GLM)$coef[,1])

but do not see how to implement confidence intervals based
on "robust error variances" using the output in GLM.

If there was a solution arrived at, can someone spell it
out for me, please?

With thanks,
Ted.

[*] PS: I did the "R Site Help and Archive Search" on
"eyestudy",
and got 18 hits, of which 6 were postings on that thread.

However, despite the fact that I requested the results to be
sorted "by date, earliest on top", they seem to come out
in somewhat random order:

Sat Jun 5 14:47:26 2004 (Frank E. Harrell)
Wed Jun 9 13:06:04 2004 (Lutz Ph. Breitling)
Sat Jun 5 13:51:45 2004 (Lutz Ph. Breitling)
Wed Jun 2 18:24:54 2004 (Frank E. Harrell)
Wed Jun 2 12:44:45 2004 (Lutz Ph. Breitling) [the initial posting]
Wed Jun 2 16:36:43 2004 (Thomas Lumley)

This made it difficult to follow the thread, and indeed I wonder
if all the postings have been found. Does Namazu have problems
in this respect? (And, if I click on "How to search", I get the
response "You don't have permission to access /namazu.html on this
server.")

And, if I click on "Contemporary messages sorted: ... by thread"
I get the response:
"The requested URL /R/Rhelp02a/archive/index.html was not found
 on this server."

--------------------------------------------------------------------
E-Mail: (Ted Harding) <Ted.Harding at manchester.ac.uk>
Fax-to-email: +44 (0)870 094 0861
Date: 08-May-08                                       Time: 14:38:36
------------------------------ XFMail ------------------------------