I'm thinking whether there exists cases where OLS estimator is closer to the true coefficient than WLS estimator in probability limit in a misspecified model.

Specifically, I'm working on a case where:

The true model consists of a dummy D, a control X and their interaction DX. But the misspecified model misses the interaction term.

  • $\begingroup$ What sort of weighting are you doing? WLS and OLS are both unbiased. WLS is just often more efficient, which isn't to do with coefficient accuracy. $\endgroup$
    – ahorn
    Commented Jan 5, 2021 at 8:03


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.