I just discovered the areg command, and I wanted to check that my use of it is correct.

If for instance I estimate the model

reg hprice bed baths garden i.district, cluster(district)

Then I estimate

areg hprice bed baths garden, absorb(district) cluster(district)

Both models give the exact same results, the only thing that is different is the degrees of freedom, which makes sense because the district fixed effect is now absorbed with areg.

There are over 1000 districts.

So, my question is, once I am not interested in estimating the coefficients on each district, I can easily substitute reg for areg instead?

  • 2
    $\begingroup$ Final question seems the wrong way round. (There is, just, a statistical question embedded here, in my view, despite the focus on Stata.) $\endgroup$ – Nick Cox Dec 14 '15 at 18:32

The degrees of freedom differ because areg makes a different assumption about what happens to the number of districts. It assumes that they stay fixed as the number of houses grows.* Clustered regress does not. Thus it seems areg is more suitable to your problem (as long as these districts are not made up by real estate agents, like Parkwanus or Rambo in New York city).

*The number of districts, however, is included in the degrees of freedom used in the finite-sample adjustment of the cluster–robust VCE estimator.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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