Page 1 of 1

Using White’s modified standard error estimates in EViews

Posted: Sun Aug 26, 2012 5:57 am
by gobbble
I have run a regression in eviews 7, which shows signs of Heteroskedasticity, according to the White test.
I have already log data, so I wish to includeWhite’s modified standard error estimates in EViews.

When I hit the option button in the estimation window, I cannot find the ‘Heteroskedasticity consistent coefficient variance’ box.

Can you please help me here? i attach a screenshot from my option menu and one screenshot with the option menu I am looking for.

Thanks a lot!

Re: Using White’s modified standard error estimates in EViews

Posted: Sun Aug 26, 2012 1:49 pm
by EViews Gareth
Looks like you're using the manual for EViews 6 with EViews 7. In EViews 7 you need to change the "Coefficient Covariance Matrix" option.

Re: Using White’s modified standard error estimates in EViews

Posted: Mon Aug 27, 2012 1:46 am
by gobbble
Alright thanks you for the hint! I selected now for the Coefficient Covariance Matrix "White". I was aware, that my coefficients would only change marginal, but is it also normal that the White test for heteroscedacity spill out the same results?

I am wondering what I can do now. All my data is already in log changes and with the option change I gues I also integrated White’s modified standard error estimates.

Any suggestions what I can do else? The problem with heteroscedacity appeared since I had to add a dummy variable to achieve normal distribution.

My regression now looks like:

ly c lx1-1 lx2 dummy1 dummy2

where l = logchange

Thanks a lot!

Re: Using White’s modified standard error estimates in EViews

Posted: Mon Aug 27, 2012 1:55 am
by gobbble
BTW, if I uncheck ‘Include White cross terms’ the results suggest no Heteroscedasticity ...

Can I just do this? When should you include it and when not?

Re: Using White’s modified standard error estimates in EViews

Posted: Mon Aug 27, 2012 10:22 am
by EViews Glenn
Your coefficient values should not change at all.

Re: Using White’s modified standard error estimates in EViews

Posted: Mon Aug 27, 2012 10:49 am
by gobbble
Sorry, my mistake. The coefficients are not changing, but the p-values are changing marginally when selecting Coefficient Covariance Matrix "White".
The Whtie test though, shows unchanged statistics and still indicates Heteroscedasticity.

Only when I uncheck ‘Include White cross terms’, my results are blue. But that is independent of selecting Coefficient Covariance Matrix "White" or not.

would appreciate your help here. Thanks!

Re: Using White’s modified standard error estimates in EViews

Posted: Wed Nov 27, 2013 12:06 pm
by haww
hye,
where can i find the "Coefficient Covariance Matrix" "White"? im having the same problem here..:'(

Re: Using White’s modified standard error estimates in EViews

Posted: Wed Nov 27, 2013 12:09 pm
by EViews Gareth
As the picture shows, on the Options page of the Equation Estimation dialog.