I am studying the determinants of patenting; accordingly my response variable in a Poisson/count data model is the number of patents in a year and country (i.e. this is a macro-panel study; I use dummies for fixed effects).

I would like to be able to compare two variables measured in different units, i.e. R&D spending in $ and the number of universities. If this were an OLS, I would simply standardize both variables (create "z-scores") and compare the effect of 1 standard deviation increase in the variables. Does it work the same way in a Poisson, i.e. are the coefficients of standardized variables in a Poisson regression output equally comparable, even though they do not have the same straightforward interpretation?

Thanks for any answer or pointer in the right direction!


1 Answer 1


When people do ths for linear regression they often also standardise the outcome variable as well which you would not do in a Poisson regression. With that caveat the interpretation would be similar.

I do wonder though with the variables you quote why you would want to do this. At present you can say something like: having one extra university is associated with a 5% increase whereas spending 10000 dollars is associated with a 6% increase. That seems more meaningful to me than saying that changing the number of universities by 0.1 standard deviation is the same as changing the spend by 0.15 standard deviations.

  • $\begingroup$ First off: Thank you very much for the clear answer, it is much appreciated! As for the your suggestions: I completely agree that the direct interpretations make more sense. I plan to present those as the main table. However, the variables I mentioned are just examples and my, um, let's call them "clients", want me to present a comparison of all variables even though that means comparing apples and oranges. A comparison in sd-terms is the best I could offer given the data. Even though I explained the caveats, they insisted that is what they want... $\endgroup$
    – asdir
    Commented Oct 18, 2016 at 14:23
  • $\begingroup$ You have our sympathies. $\endgroup$
    – mdewey
    Commented Oct 18, 2016 at 15:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.