Heteroskedasticity

Using generalized least squares and calculating robust standard errors are possible remedies for heteroskedasticity phewww Schweser doesnt saya thing abt generalised least squares ??? does it

Yes it does

Book 1, page 239

I thought you use the “White” test…is that the same thing as adjusting the SE of the coefficients?

test hetero using Busch Pagan and use White corrected standard errors to correct

Right, so the White method is correcting the SE by INFLATING IT right? Do you think we need to know the details of the White (correct for Hetero) and Hansen (correct for Serial correlation)?

Know hansen corrects for serial correlation and heteroskedactisity at the same time, but don’t use hansen’s method if you only have conditional heterskedactisity.

yes, white SE would be larger, therefore bringing down the t-stat and eliminating too many type I errors.

here’s my pseudo langauge… Hetero-Test: BP Hetero-Correct: White/ GLS / RSE SC-Test-non-AR: DW SC-Test-AR: t-stat sgnificance SC-Correct: Hansen (Bonus: Corrects Hetero too)

if residuals have an ARCH process then the correct remedy is GLS which allows us to better interpret results