correcting t stat in panel

[obicna89]

hello!

i have a panel data analysis. least squares estimation, fixed effects.
so, i want to correct t statistics for heteroscedasticity, but i don't know how.
what do i have to do? do i go to panel options and change weights and/or coefficient covariance method?

[obicna89]

please help me if anyone knows :(

[Econometrics_is_fun]

please help me if anyone knows :(

Hello obicna(=ordinary)!

First, in panel data sets you not only will have problems with heteroscedasticity but autocorrelation as well.
I would thus advise you to select White Cross section in coefficient covariance method menu.

[obicna89]

hello. thank you!
so, by selecting white cross section i can eliminate both problems (autocorr and heteros.)?

[startz]

No.

[obicna89]

so how do i correct autocorr?

[Econometrics_is_fun]

so how do i correct autocorr?

You correct for H and AC as I wrote earlier.
With this method you can mitigate the problems facing you but not remove them. Autocorrelation in panel data is truly an enemy to us all. You can read more about this here http://forums.eviews.com/viewtopic.php?f=4&t=4448

[obicna89]

many thanks!
:)

[startz]

The White cross section method corrects standard errors for heteroskedasticity. It does not correct for autocorrelation.

[obicna89]

The White cross section method corrects standard errors for heteroskedasticity. It does not correct for autocorrelation.

when you said no, i asked you how can i correct autocorr, i didn't ask you this (quoted).

[startz]

I was responding to the post by Econometrics_is_fun.

I don't believe that EViews offers autocorrelation consistent standard errors in panels. You can, however, include an AR(1) term in the regression.

[obicna89]

I was responding to the post by Econometrics_is_fun.

I don't believe that EViews offers autocorrelation consistent standard errors in panels. You can, however, include an AR(1) term in the regression.

oh, i am sorry. i thought you were respondig to mine comments :)
the problem is, when i use ar(1) , eviews says near singular matrix :?

[startz]

That's puzzling.