Negative binomial regression vs poisson regression in panel data

[Christopher Weber]

Hello everyone,

is there a (simple) way to test my panel data set to see if I should use a negative binomial model or a poisson model to conduct the regression?

best regards

Chris

[Jared Greathouse]

Check for overdispersion. https://is.gd/3opkbA

[Christopher Weber]

Hello and thank you for your reply! But how do you do that in Stata?

best regards

[Jared Greathouse]

I'm not in front of my computer, but nbreg does this itself. If it exists, you can use gnbreg, or an overdispersed poisson model.

Or, you can use the overdisp command.

[Christopher Weber]

Thank you Jared, you helped me a lot.

Best regards and best wishes

[Christopher Weber]

One additional question, if I may: Can overdispersion be too high even for negative binomial regression models? While the mean of my dependent variable is 113 and my variance is 188478...

best regards

[Fei Wang]

I'm not intentionally downplaying nbreg, but literature shows that you should use xtpoisson rather than xtnbreg, regardless of overdispersion. See discussion here.

[Christopher Weber]

Thank you for the reply, Fei Wang!