Re: st: Compare fit of variable in modified poisson model

[Maarten Buis <maartenlbuis@gmail.com>]

On Tue, Dec 17, 2013 at 3:34 PM, Kassie Melius wrote:
> I am using Stata 13 and modified poisson regression with robust error
> variances (as per Zou G. A Modified Poisson Regression Approach to
> Prospective Studies with Binary Data. Am J Epidemiol 2004;
> 159(7):702-6.)
>
> How can I compare the p-value for the fit of each variable in my model ?
>
> For example if i run:
>
> glm depvar i.var1 i.var2 i.var3 i.var4, fam(poisson) link(log) nolog
> vce(robust) eform
>
> Is it appropriate to use testparm for each var1-var3 given that it is
> a Wald test, or should I be using lrtest (or something else)?

In principle you could use either the Wald test or the likelihood
ratio test, they are asymptotically equivalent. The likelihood ratio
test tends to work slightly better in small samples, but in practice
the number of observations is either going to be too small for both
tests, or so large that both tests work fine. The window in which the
likelihood ratio test works well and the Wald test does not is very
small. So I would not spent a single second of your valuable time on
this issue, just use whichever is most convenient.

However, in your case this entire discussion is useless: you cannot
use a likelihood ratio test and your only option is the Wald test
since you used the -vce(robust)- option.

Hope this helps,
Maarten

---------------------------------
Maarten L. Buis
WZB
Reichpietschufer 50
10785 Berlin
Germany

http://www.maartenbuis.nl
---------------------------------
*
*   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/