Bookmark and Share

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

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

Re: st: Poisson Regression

From   Maarten buis <>
Subject   Re: st: Poisson Regression
Date   Mon, 14 Feb 2011 11:50:21 +0000 (GMT)

--- On Mon, 14/2/11, Alexandra Boing wrote:
> I dont understand,"curve still look resonable" 

Remember that -poisson- for a binary variable is a model
that cannot be true. Under certain circumstances it
can still be useful. You need to explore the 
consequences of your model on your data to see if
you are just pushing your model/data too hard or if
it is still reasonable. This requires that you know
what would be reasonable outcomes for your research
topic, but you should have ideas on that anyhow.

To be concrete just execute the example as I sent it
and than just look at the graph. If that looks 
reasonable to you then it looks reasonable. It is as
simple as that. There is no magic involved in these
types of model checking, it is just knowledge on the 
process being studied that is being compared with the 
outcomes of your model.

> and "curve from a -logit- regression, which would
> be the obvious alternative when -poisson- leads
> to unrealistic predictions".

Again, execute the example as I sent it. You will
see two curves: one from -poisson- and one from
-logit-. The curves created by -logit- will always
remain within the [0, 1] range, and are for that
reason the logical alternative. If you choose 
-logit- than you will than need to give up on 
interpreting the results as risk ratios and instead 
move to odds ratios.

Hope this helps,

Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen


*   For searches and help try:

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