Page 1 of 1

Seemingly Unrelated Regression

Posted: Wed Mar 14, 2012 6:34 am
by canboardrp
Hello,

I'm trying to estimate the following using the seemingly unrelated equations methodology but I get an error, near singular matrix prompt. What's the cause of this and how can I adjust the programming to ensure that this doesn't occur?
LOG(RTDDPE) = C(1) + C(2)*LOG(REPNL)+C(3)*LOG(POPNL)+C(4)*LOG(HDDNL)
LOG(RTDDPE) = C(5) + C(6)*LOG(REPPE)+C(7)*LOG(POPPE)+C(8)*LOG(HDDPE)
LOG(RTDDNS) = C(9) + C(10)*LOG(REPNS)+C(11)*LOG(POPNS)+C(12)*LOG(HDDNS)
LOG(RTDDNB) = C(13) + C(14)*LOG(REPNB)+C(15)*LOG(POPNB)+C(16)*LOG(HDDNB)

Re: Seemingly Unrelated Regression

Posted: Wed Mar 14, 2012 6:47 am
by startz
What happens when you estimate the equations separately by LS?

Re: Seemingly Unrelated Regression

Posted: Wed Mar 14, 2012 6:49 am
by canboardrp
The equations when estimated using LS work.

Seemingly Unrelated Regression

Posted: Wed Mar 14, 2012 7:02 am
by EViews Gareth
Don't you have the same dependent variable twice?

Re: Seemingly Unrelated Regression

Posted: Wed Mar 14, 2012 7:04 am
by canboardrp
Gah. I'd missed that typo in the programming.

Thanks.

Re: Seemingly Unrelated Regression

Posted: Tue Jul 22, 2014 5:29 am
by econworker
Hi Eviews Gareth, I hope you reply my question here.
I am trying to set up this model in Eviews to estimate it with using a SUR system model:

RHO_GOLD_GASOLINE = C(1) + C(2)*LOGD_TBILLS + C(3)*LOGD_USEX + C(4)*MAC_ADS + C(5)*RHO_GOLD_GASOLINE(-1)+ [AR(1)=C(6)]
RHO_GOLD_HO = C(7) + C(8)*LOGD_TBILLS + C(9)*LOGD_USEX + C(10)*MAC_ADS + C(11)*RHO_GOLD_HO(-1) + [AR(1)=C(12)]
RHO_GOLD_NG = C(13) + C(14)*LOGD_TBILLS + C(15)*LOGD_USEX + C(16)*MAC_ADS + C(17)*RHO_GOLD_NG(-1) + [AR(1)=C(18)]
RHO_GOLD_WTI = C(19) + C(20)*LOGD_TBILLS + C(21)*LOGD_USEX + C(22)*MAC_ADS + C(23)*RHO_GOLD_WTI(-1) + [AR(1)=C(24)]

The problem is that if I estimate the system without AR(1) term there is no problem and it works fine, however when I include the AR(1) term to the model and I push ok to estimate it, Eviews suddenly stop to work and I have to close it completely, I tried it with different sets of data and still getting the same problem. I would like to know do you have any idea why Eviews can not estimate a SUR model that includes AR(1) term and lagged dependent variable in the same time? can it be a problem with my Eviews?
If you need I can attach you the workfile.
Thanks

Re: Seemingly Unrelated Regression

Posted: Tue Jul 22, 2014 6:50 am
by startz
The first thing Gareth is going to ask you is what the version and build date is of your EViews copy. Look under Help/About.

Re: Seemingly Unrelated Regression

Posted: Tue Jul 22, 2014 7:32 am
by econworker
The first thing Gareth is going to ask you is what the version and build date is of your EViews copy. Look under Help/About.
Thanks for quick reply Startz, its Eviews 8, 19 November 2013 build.

Re: Seemingly Unrelated Regression

Posted: Tue Jul 22, 2014 7:33 am
by startz
You might just do an update and see if that happens to fix the problem.

Re: Seemingly Unrelated Regression

Posted: Tue Jul 22, 2014 8:39 am
by econworker
You might just do an update and see if that happens to fix the problem.
It is solved thanks alot.