Heteroscedasticity-robust F-statistic

[Anthony]

I am estimating a time-fixed effects model. The white test suggests the presence of heteroscedasticity. I have corrected for this using the White period robust coefficient variance estimator. When I apply this correction, the coefficient standard errors change, and in turn the t-statistic and prob(t-statistic) also change.

I would expect prob(F-statistic) to also change, but it does not. I say this because when estimating using one determinant, prob(t-statistic) = probability(F-statistic). Therefore if the standard errors change after applying the robust estimator, prob(t-statistic) changes and prob(F-statistic) should also change. This implies that Eviews does not report heteroscedasticity robust F-statistics. Could you please guide me how to go about this problem in Eviews.

Thanks,

Anthony

[EViews Gareth]

If you perform a Wald test (View->Coefficient Tests->Wald Test), it will use the robust standard errors.

[Anthony]

Thanks for this. The wald test works. What criteria do you use to select the order of polynomial to use in the regression to be tested? For example, when do you use yhat^2, and whnet do you use yhat^2, yhat^3 etc?