0

I am estimating a model based on time series, which comes from theoretical background (economic theory), and the specification is quite common in empirical literature. However, I find that estimated errors are correlated, and therefore Breusch-Godfrey LM test does not reject null hypothesis of no autocorrelation. I am inclined do not change specification of the model (as I mentioned it comes from ecnomic theory), but autocorrelation bothers me. Since under autocorrelation, estimated coefficients are still unbiased (but variances of coefficients are biased), so can I simply neglect the autocorrelation? If no, what can I do without changing specification? I think to use HAC (Newey West) estimator to report in my regression table robust standard errors. Thank you in advance!

Richard Hardy
  • 54,375
  • 10
  • 95
  • 219
sane
  • 131
  • 3
  • 1
    Check out [this](https://stats.stackexchange.com/questions/181257/correcting-for-autocorrelation-in-simple-linear-regressions-in-r/181297#181297), [this](https://stats.stackexchange.com/questions/226279/how-can-i-handle-autocorrelated-residuals/226441#226441) and [this](https://stats.stackexchange.com/questions/191819/hac-standard-error-or-missing-arma-terms/191830#191830). – Richard Hardy Nov 28 '19 at 10:07
  • 1
    Thanks for the useful links! – sane Nov 28 '19 at 10:28
  • You are welcome! You may also consider whether your question might be a duplicate of any of those I linked to. Otherwise, you may indicate what is left uncovered in your question. – Richard Hardy Nov 28 '19 at 10:47
  • I think that this question can be considered as a duplicate of one of them. But I don't know how to attach it to them. – sane Nov 29 '19 at 15:44

0 Answers0