3

Since my residuals failed the test for heteroscedasticity and autocorrelation, what is the benefit of using Newey-West when there is no impact on the coefficient estimates in OLS?
Does the Newey-West make any impact on the OLS model?
Should I use GLS instead of OLS due to heteroscedasticity and autocorrelation?

Richard Hardy
  • 67,272
StatsUser
  • 1,809
  • 1
    Here are some related threads. – Richard Hardy Nov 26 '22 at 16:20
  • @RichardHardy, since Newey-West, produces alternate standard errors as an alternative to the regular standard errors, there is no change in the coefficient. so the model is misspecified; less accurate forecasting due to neglecting the autocorrelation in model errors. Does the Newey-West make any impact on the OLS model? – StatsUser Nov 26 '22 at 16:36
  • How to interpret when the P-value of any predictors using the Newey-west test is not significant. – StatsUser Nov 26 '22 at 16:40
  • 1
    OK, here is a more concrete reference. Newey-West keeps coefficient estimates intact but alters the standard errors. The interpretation of a $p$-value above the significance level $\alpha$ is as usual. – Richard Hardy Nov 26 '22 at 16:54

1 Answers1

1

If you only want a predictive model where you take the coefficient estimates and apply that regression equation to new data, the Newey-West does nothing for you.

However, Newey-West is a pretty good way of handling unequal variance and autocorrelation in the error terms when it comes to estimating standard errors for the regression coefficients.

If you are interested in testing or writing confidence intervals around regression coefficients, Newey-West could be a valuable tool. Newey-West was first published in an econometrics journal, and much of economics deals with this sort of inference rather than pure predictive modeling.

Dave
  • 62,186