Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: RE: RE: Using command svy glm to obtain risk ratios


From   Steve Samuels <sjsamuels@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: RE: RE: Using command svy glm to obtain risk ratios
Date   Mon, 17 Feb 2014 20:40:39 -0500

Amena:

I recently co-authored a paper for which I first ran 
-svy: poisson-

and then 
-svy: glm, link(log)

which failed in about 2 of 20 models.

I found four or five instances in which the relative risks differed so much that we had to abandon the idea of the Poisson approach. In others, the CIs for the RR exceeded 1.0  We were able to get -svy: glm, link(log) to work by slight modification of the models.

You still have not shown us the actual commands you ran and the problematic -svy: glm- output.
Please do so.  Right now we have no actual evidence of the problem.. The only difference that I can see in the output of -glm- and -svy: glm, with the "eform" option is that -glm- heads the  coefficient table "Risk Ratio" whereas -svy: glm- has a heading "exp(b)", which is the same thing.


Steve Samuels
sjsamuels@gmail.com



On Feb 17, 2014, at 6:30 PM, "Hussein, Mustafa (Mustafa Hussien)" <mhussei4@uthsc.edu> wrote:

You are welcome! 

No, you should estimate your models with the -svy- prefix if you have complex survey data. Your code should look like:

svy: glm depvar indepvar, fam(poisson) link(log) eform

Using the -svy- prefix is equivalent to cluster robust estimation with pweights. So, in essence, the "robust" part of modified Poisson is taken care of when using the -svy- prefix. 

For more info on estimating model-based RRs using log-binomial, Poisson, and Cox models, see:
Barros & Hirakata (2003). BMC Med Res Methodol 3(21). PMID: 14567763
Zou (2004). Am J Epidemiol 159:702–706. PMID:15033648

Hope that helps
Mustafa 
________________________________________
From: owner-statalist@hsphsun2.harvard.edu [owner-statalist@hsphsun2.harvard.edu] on behalf of Agunwamba, Amenah A., Ph.D. [Agunwamba.Amenah@mayo.edu]
Sent: Monday, February 17, 2014 4:17 PM
To: statalist@hsphsun2.harvard.edu
Subject: st: RE: Using command svy glm to obtain risk ratios

Thanks Mustafa,

So if I just eliminated the "svy" command, I can still get weighted RRs?

An example code might be:

glm outcome predictor, fam(poisson) link(log) no log vce(robust)

Thanks!


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index