Specifying fixed effects in an OLS model

BeeBop9 — Wed, 03 Apr 2019 19:52:44 GMT

I want to build a regression model (OLS) with the following:

My data set is at the medical provider level (N=10,000 approx). I have several providers within each US state, sometimes hundreds.

My dependent variables is "hours worked (for various categories of worker) per patient". This is at the provider level.

My independent variables are: state (I have 52 states), and the minimum wage in each state (state level variable).

I want state to be a fixed effect, since there are many providers in each state.

How do I specify this model --PROC GLM? Or is there another proc that's more appropriate?

Thanks!

-Dave

Re: Specifying fixed effects in an OLS model

PaigeMiller — Thu, 04 Apr 2019 13:19:22 GMT

Yes, you can do this in PROC GLM.

If you don't specifically designate variable STATE to be random, then it is assumed to be fixed.

I want state to be a fixed effect, since there are many providers in each state.

As far as I know, this is not a reason for variable STATE to be fixed. (It should be fixed, but not for that reason).