Equality of Distributions across nominal categories?

[Peter Meier]

Dear all,

I would like to test the null hypothesis that observations from 2 samples S1 and S2 are equally distributed across the 50 states. My best guess for testing this would be -ksmirnov- but I am a bit unsure whether this works also when the 50 categories are 50 different states for which I have no sensible way of ordering them? Or would you suggest a more appropriate test?

Thanks so much and best regards,
PM

[Nick Cox]

Sounds like chi-square to me. If you mean the 50 states of the US, it's hard to understand why that null is of interest.

[Peter Meier]

Thanks!

I have a sample with observations across all states and would like to test to what extent regression results are representative of the entire country.

[Nick Cox]

For that you need independent information, I think.