Heteroskedasticity

[sextonmb]

Hi,

I am in an econometrics class this semester, and we are working a lot with finding and fixing heteroskedasticity in regression models. I was wondering how to fix the coefficients once I run the White test and find there is heteroskedasticity. Any help would be greatly appreciated.

Thanks

[startz]

Basically you need to build a model of the variances of the error terms and then use weighted least squares.

[sextonmb]

If I have 3 variables the model would be 1/square root(x1*x2*x3), correct? Also, I was wondering what the Newey-West output corrects because I believe my professor mentioned that the Newey West model automatically fixes something I'm just not sure if it is the coefficients. I really appreciate the help though.

Thanks

[startz]

No, there isn't any one magic way to fix heteroskedasticity. You actually have to build a model for the variance.

Newey_West corrects standard errors for both heteroskedasticity and serial correlation.

[sextonmb]

Alright, but to build the homoscedastic model, if I multiply the original regression by the 1/sqr(variables) weight, will that give me coefficients that are BLUE? And again, I truly appreciate the help.

[startz]

If you multiply by 1/sqr(variance), then you should be in good shape.