Linearity of the logit- Poisson Regression

[ValSki]

Hi friends,

I'm doing a preliminary investigation of my data and would like to test the linearity of the logit for my age variable (continuous), but I'm at a loss. I'll be running a Poisson regression. For binary logistic regression, I'd typically run a Box-Tidwell test, but unsure on what the best approach is for a Poisson and/or if it's even different. It's my first time investigating data with a Poisson distribution.

Any help on the method and the syntax for SAS would be greatly appreciated. Thanks in advance!

[StatDave]

There is no logit involved in a Poisson regression model. The logit is only involved in logistic models. However, I suspect that what you actually mean is that you want to test for having the proper functional form of your predictor in your Poisson model or that the log link is appropriate. See the "Linearity in the logit (or link), testing" item the list of Frequently Asked-for Statistics (FASTats) in the Important Links section of the Statistical Procedures Community page. As noted there, you can use the ASSESS statement in PROC GENMOD when fitting your Poisson model to assess both the functional form of the predictor and the link function. Also as noted, using a Generalized Additive Model model, such as a model that uses a spline for your predictor, can help you find a suitable representation of the predictor in your model. 

[ValSki]

Great, thank you!

[sbxkoenk]

Hello,

Just responding to this part of the title : "Linearity of the logit"

SAS Tip: Test assumptions of linearity with %logitplot

https://communities.sas.com/t5/SAS-Tips-from-the-Community/SAS-Tip-Test-assumptions-of-linearity-wit...

Cheers,

Koen