white test after weighted least squares

[russellrobins]

this question is about how EVIEWS performs the WHITE test for heteroskedasticity AFTER you have estimated the model by weighted least squares

i estimate (1) y = b0+b1*X +b2*Z + e....build a model for heteroskedasticity log(e_hat) = a0+a1*x+a2*Z.....forecast e_hat (call the forecasts e_hatf) then estimate 1 again as y/e_hatf^.5 1/e_hatf^.5 x/e_hatf z/^hatf......then test for heteroskedasticity (using white again)

the white test after the y/e_hatf^.5 regression is NOT the same as the white test if i lets EVIEWS estimate wls (using weights (inverse standard deviation..etc--although the coefficients and standard errors are the same).

i must be missing something. can you point out my eviews error or perhaps give me a cite for a paper that shows what EVIEWS is doing for the white test AFTER estimating wls.

thanks

Russ

[EViews Gareth]

The White test is a residual based test. When you estimate with WLS, EViews uses the unweighted residuals.

When you estimate manually you are forcing use of the weighted residuals.