Thank you for both explanations. I assume that your point about an IR above 1
still holds if exposure is a number of attempts (and the numerator is the
number of successes), correct? I ask this because that is my situation and
here an IR above 1 means a predicted value of successes that is greater than
the number of trials (the exposure).
Are there sensible constraints that I could use on the model to keep the
predicted values to no more than the exposure? Are there any alternative
models that I should look into?
Thanks,
Rich
German Rodriguez wrote:
Joe,
I believe everything depends on whether your Poisson model includes an
offset. This is explained in the help system if you type
help poisson postestimation, marker(predict)
If the model doesn't have an offset, then |predict ir, ir| gives the fitted
count. This is the same as |predict mu| (with no options), and the same as
|gen mu = exp(xb)| after |predict xb, xb|.
If the model has an offset, then |predict ir, ir| gives the predicted rate,
which is the number of events divided by exposure. This is not the same as
|predict mu|, which gives the expected count taking exposure into account.
Note also that |predict xb, xb| gives the linear predictor *including* the
offset. Another way to obtain the incidence rate is |gen rate =
events/exposure| after |predict events|, assuming your exposure variable is
called exposure.
In either case it is possible for the rate to exceed one; you could expect
more than one event (no offset) or more than one event per unit of exposure
(with offset). For example the number of children ever born per woman can be
more than one.
Cheers,
German
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
