Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

RE: st: RRR with CI from logit model


From   Leonelo Bautista <lebautista@wisc.edu>
To   statalist@hsphsun2.harvard.edu
Subject   RE: st: RRR with CI from logit model
Date   Tue, 02 Nov 2004 10:16:58 -0600

Jose Maria is right. -logistic- converges in cases where -binreg- and -glm-
fam(bin) link(log) do not. However, in cases where the odds ratio is not a
good estimate of the relative risk, it is worth trying -binreg-. 

Now that you bring up this issue, years ago I asked why the first -binreg-
(from STB50) converged in cases where the -binreg- that came with Stata 7.0
did not. Bobby Gutierrez was kind to explain that, in Stata 7, -binreg- was
"just a fancy front end to -glm, fam(binomial)-", and that in cases of
numerical instability slight differences in algorithms between -binreg- and
-glm- would produce different answers. Now, Bobby's answer made me suspect
that the corrections to keep the predicted probabilities within the 0-1
range, which were implemented in the first -binreg-, may not have been
preserved in future versions. As far as I understand, -binreg- from Stata
8.0 is the same as -binreg- from Stata 7.0. I'd be relief if somebody
confirms that my suspicion is unfounded.

A few months ago, I tried the first -binreg- and the one in Stata 8.0 in the
same large data set and found the same problem: using the rr option, the
original -binreg- converges, but the new one does not. My question elicited
no response. 

Just in case, I still keep and prefer to use the first version of -binreg-,
which I verified was consistent with the GLIM code used by S. Wacholder when
he first proposed using binomial regression (Am J Epidemiol. 1986
Jan;123(1):174-84).


Leonelo Bautista

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of jmpsouza@usp.br
Sent: Tuesday, November 02, 2004 9:18 AM
To: statalist@hsphsun2.harvard.edu
Subject: RE: st: RRR with CI from logit model

Citando Leonelo Bautista <lebautista@wisc.edu>:

> I think Ronan Conroy's advice in this case is very sound. .....> 
> On the other hand, I fail to see why you want to calculate relative risks
> from a logistic model.  If that's the case, as Joseph
> Coveney suggested, you can use -glm- fam(bin) link(log), or -binreg-. That
> should allow you to calculate absolute and relative risks with their
> confidence intervals for any values of your independent variable. 
> ....
> Leonelo Bautista
 
Dear Statalisters:
let me enter late at this thread, to remind that -binreg- or -glm- f(bin)
link(log) many times fail to converge, in situations when logistic runs
well.
Cheers,
Jose Maria

Faculdade de Saude Publica/ Departamento de Epidemiologia
Sao Paulo

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index