Correct for Autocorrelation

[lidis]

Hi,

My data: I have panel data. Each individual rated 5 brands on 1 attribute and overall satisfaction. I want to pool across brands and individuals and estimate via OLS.
(My regression function is: AttributeRating= AverageAttributeRating + OverallSatisfactionRating
Since each individual rated each of the 5 brands, my errors are not IID.

So I need to correct for autocorrelation&heteroskedasticity.

The latter one is easy (I just tick the box), but how do I correct for autocorrelation?
I read so many different approaches (eg. AR(1)) and don't know what is the right solution for me.

Thank you in advance for your advice!

[lidis]

When I just choose the to estimate robust standard errors (hac) with newey-west in eviews7 my durbin watson statistic does not change...

Can't anybody give me a hint? I really have to get this done by today

[startz]

Newey-Watson doesn't change the regression, just the standard errors. That's why the Durbin-Watson doesn't change.

[lidis]

Thank you for the reply :D

I thought Durbin Watson indicates autocorrelation?

So is it enough for me just to use newey west?

For more details:
My data is in the following form & and I want to run a linear regression pooled across individuals and brands.
person performancerating brand
------ ---------------- -------
1 2 samsung
1 3 nokia
1 2 iphone
2 1 samsung
2 5 nokia
2 5 iphone
3 2 samsung
3 4 nokia
3 4 iphone
. . .
. . .
. . .

[startz]

Durbin-Watson does indicate serial correlation. Newey-West corrects standard errors to account for serial correlation.