WLS method

[Duong Chi]

Dear Statalist,
I have a project with cross-sectional data, but my data have problem with heteroscedasticity. WLS is the method which was used to solve this problem. Using regwls in Stata, but I do not know (1): wvar: how we can determine varlist to be weighted. (2) which type is the best (e2; loge2; xb2 or abse)

[Carlo Lazzaro]

Duong:
welcome to this forum.
Have you already taken a look at https://stats.idre.ucla.edu/stata/ad...es-regression/ ?
That said, I think you would be more comfortable with invoking -robust- statndard errors in -regress- in order to take heteroskedasticity into account.

[Duong Chi]

Dear Prof Carlo,
Thanks for helping me

[Deboshmita Brahma]

Carlo Lazzaro Sir, I have done robust standard errors while doing OLS. However, when I predict the residual after the OLS regression, then generate the residual square and regress it on the explanatory variables, most of the coefficients are significant. Does this imply my model still has the problem of heteroscedasticity? I would be grateful if you please help, Sir.

[Carlo Lazzaro]

Deboshmita:
plese share what you typed and what Stata gave you back (as per FAQ). Thanks.