panel estimation, robust least squares, heteroskedasticity

[Wouter van der Stee]

Dear,

I have two questions:

1): I'm currently running a regression with panel data. In order to put less weight on outliers I would like to run a robust least squares regression.
This function is however not incorporated in the version (7) I use. Is there a way to do something similar? For now I have deleted the outliers based on the residual table, is this a decent way?

2): Is there a function to check for heteroskedasticity in my panel of data? When I read the data as unstructured there exists heteroskedasticity. Can I then just simply choose for the white-cross section in the coefficient covariance menu? Or does the heteroskedasticity in the unstructured data does not necessarily mean heteroskedasticity exists also in the panel of data?

Sincerely,

Wouter van der Stee

[EViews Gareth]

You'll need to upgrade to EViews 8 for robust least squares.

There are no panel tests for heteroskedasticity built in.

[Wouter van der Stee]

So is it then a sound move to use 'white cross section' in the coefficient covariance method menu for my panel data because I found heteroskedasticity in the unstructured data set (same data as panel)? Or should I just select ordinary ?