R squared - Negative binomial

[Jack Smith]

Hello everyone!



I am currently performing some negative binomial regressions (xtnreg) with my panel data.
I tried to measure the R squared but it seems that it is not really common when using a negative binomial regression. From what I saw, I should measure the Maximum Likelihood estimator.
Here is the source: http://cameron.econ.ucdavis.edu/rese...96preprint.pdf

My question is why is it the case? Why can't we use R^2? And what would be a measurement that could replace it?


Thank you
-Jack

[Marcos Almeida]

Maybe this thread and this FAQ will be helpful to you.

[Jack Smith]

Hi, thanks for the reply!

I understand that I can get the McFadden’s pseudo R-squared with xtpoisson.
However, I do not understand why I can't even get a Pseudo R-squared with a negative binomial function..
I've been looking around and did not find any information about it.

Would you please explain me why it is so?

Thank you
-Jack