Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: suest across two svy:glm models to test interaction


From   Steven Samuels <sjhsamuels@earthlink.net>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: suest across two svy:glm models to test interaction
Date   Mon, 8 Sep 2008 19:34:05 -0400

Christy-


Your -svy- statements are not correct: binge2lt = 0 for those who binge ³ 1 month, and so includes those people in the reference group; similarly, binge2gt will include those the <1 month people in the reference group. You need -subpop- to include only the two groups in which you are interested.

Like, Maarten I think that -suest- is a bad idea. I too think that your "common outcome" logic for using the RR model is faulty. But I would go further: with common outcomes, main effects, as well as interactions, are hard to interpret, on the OR scale and the RR scale. I strongly suggest that you convert to the probability scale to have understandable results. That will not be easy to do with - suest-, for there is no need for the two "unrelated" equations to be compatible (i.e. to lead to three probabilities that sum to 1.

Another difficulty: with a Poisson family assumption, your relative risk models can easily lead to estimated probabilities, or their CI upper-endpoints, which are >1. A command that fits the probabilities for all three outcomes is much better. Before running one of the ordered logit commands, as Maarten suggests, try -svy: mlogit-. After these commands, run -mfx- or -margeff- (download from SSC) to show estimated effects on the probability scale.

-Steve

On Sep 8, 2008, at 5:20 PM, Christy McKinney wrote:


I am examining the association between binge drinking and neighborhood
poverty using survey data. Because binge drinking is highly prevalent
I
would like to use glm to estimates relative risks (instead of logistic
regression
& odds ratios).

Past year binge drinking is categorized into 3 groups: no binge
(referent); binge <1 month; and binge ³1 month. In one glm model, I
compare binge <1 month (binge2lt) to no binge drinking. In a separate
glm model, I compare binge ³1 month (binge2gt) to no binge drinking.

I want to evaluate whether the risk of binge drinking associated with
neighborhood poverty is different for men and women (e.g. does sex
modify the relation between binge drinking and neighborhood poverty?).
I
am new to the suest command and am not sure I am using properly. Can I
use the suest command to combine the two separate models and test the
interaction jointly across binge drinking categories?

Below is some example code of what IÕm trying to do. Is this
correct?
Thanks in advance for any assistance! -Christy

EXAMPLE CODE:

xi: svy: glm binge2lt poverty sex i.sex*poverty, fam(poisson)
link(log)
nolog eform
	est store b1
xi: svy: glm binge2gt poverty sex i.sex*poverty, fam(poisson)
link(log)
nolog eform
	est store b1a
suest b1 b1a
test [b1_binge2lt]_IsexXpover_1 [b1a_binge2gt]_IsexXpover_1

Christy McKinney, PhD, MPH
Faculty Associate, Epidemiology
UT Houston School of Public Health
Dallas Regional Campus
6011 Harry Hines Blvd., V8.106
Dallas, TX 75390-9128
christy.mckinney@utsouthwestern.edu
(214)648-6574 (phone)
(214)648-1081 (fax)


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/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/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