how to remove 1st order autocorrelation in unbalanced panel

For econometric discussions not necessarily related to EViews.

Moderators: EViews Gareth, EViews Moderator

dancker
Posts: 4
Joined: Thu Feb 27, 2014 3:29 am

how to remove 1st order autocorrelation in unbalanced panel

Postby dancker » Thu Feb 27, 2014 3:42 am

Hi,

I'm having troubles solving the following. Please bear in mind that my knowledge of econometrics and eviews 7 is fairly limited..

My panel data set, in which I regress several explanatory variables on the natural logarithm of price to net asset value (monthly basis) of 25 UK investment trusts over a 10 year period, is suffering from first order autocorrelation (DW statistic is 0.48). So I have monthly premium levels of 25 different UK investment trusts over a 10 year period, that I try to explain through several (8 different ones) explanatory variables. The tip I got from my university was to use Newey-West standard error's but I've already read in another post that this is not possible in Eviews for panel data. Hence, how can I remove this autocorrelation (and actually also heteroskedasticity) from my model?

Hope that someone here knows what to do..

ps. I'm probably forgetting relevant info to come up with a meaningful answer so let me know if any further info is required.

Thank you in advance.

dancker
Posts: 4
Joined: Thu Feb 27, 2014 3:29 am

Re: how to remove 1st order autocorrelation in unbalanced pa

Postby dancker » Thu Feb 27, 2014 7:37 am

edit: I've found out that in Eviews: in the 'equeation estimation' menu --> 'panel options' tab, i can choose 'Weights' and 'Coef covariance method'. Can anyone advise me what to choose here to overcome the first order autocorrelation in my dependent variable?
My dependent variable is a monthly share discount or premium to the net asset value per share of the investment trust. This discount or premium is heavily autocorrelated with the discount/premium of the month before.

I'd appreciate the help!

EViews Glenn
EViews Developer
Posts: 2682
Joined: Wed Oct 15, 2008 9:17 am

Re: how to remove 1st order autocorrelation in unbalanced pa

Postby EViews Glenn » Thu Feb 27, 2014 11:22 am

You can add lagged dependent regressors to whiten the residuals, but with fixed effects you'll probably have to use a GMM approach. See the literature on dynamic panel data.

dancker
Posts: 4
Joined: Thu Feb 27, 2014 3:29 am

Re: how to remove 1st order autocorrelation in unbalanced pa

Postby dancker » Mon Mar 31, 2014 1:53 pm

Dear Glenn, thank you for your response.

Another question regarding this topic; how do i diagnose if my data suffers from heteroskedasticity? As my data is panel data I am not sure how to check if and where there is heteroskedasticity and how to adjust for it. Can someone help me out?


Return to “Econometric Discussions”

Who is online

Users browsing this forum: No registered users and 2 guests