Re: st: RRR with CI from logit model

[Ron�n Conroy <rconroy@rcsi.ie>]

jmpsouza@usp.br wrote:

> 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
>
>  
Jose Maria and I have different experiences of -binreg-. This week I 
have run a whole bunch of regressions that did, indeed, converge, and I 
reckon that more than 80% of the models I've tried have converged.
One strategy to getting -binreg- to run is to avoid adding all potential 
confounders to the model. Use -logistic- to identify important 
confounders and then run -binreg- on the resulting model. A few -binreg- 
models initially failed to converge for me, but I realised that the 
person who had done the original analysis had added a number of 
unimportant predictor variables. Once these were pruned, the stuff ran.

Relative risks make so much more sense than odds ratios (at least to me 
- I'm just a keyboard player) that I always try -binreg- in situations 
where the data represent incidences.

--

Ronan M Conroy (rconroy@rcsi.ie) 
Senior Lecturer in Biostatistics 
Royal College of Surgeons 
Dublin 2, Ireland 
+353 1 402 2431 (fax 2764) 
-------------------- 
Just say no to drug reps 
http://www.nofreelunch.org/

*
*   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/