Page 1 of 1

3sls / joint estimation: residuals

Posted: Fri Mar 09, 2012 4:45 am
by worldboy
Dear all,

Is it possible in EViews (7) in a system to grab the residuals from one of the equations and use them in another (equation of that system)? (and how?)

I have a rational expectaions model in which I want to use the one step ahead prediction errors of one equation (=its residuals from one period ahead) as regressors in one of the other equations. Something like:

(eq1) Ft = a + bXt+ ERRORt

(eq2) Zt = c + ... + dERRORt+1 + ...

I hope that's clear enough...

Thanks and best regards

Re: 3sls / joint estimation: residuals

Posted: Fri Mar 09, 2012 8:48 am
by EViews Gareth
No, that's not possible in a system. Of course if the only interaction between the two equations is through that residual series, you could always estimate them as separate equations. I presume that your choice of 3SLS means this is not the case though :(

Re: 3sls / joint estimation: residuals

Posted: Fri Mar 09, 2012 9:39 am
by worldboy
Thanks Gareth. Your presumtion is right. I have a joint determination of coefficients in the system. Well, it's unfortunate that Eviews does not allow something like this. Maybe in a future version?
:roll:
CHR

Re: 3sls / joint estimation: residuals

Posted: Fri Mar 09, 2012 10:38 am
by EViews Gareth
You could manually put the residuals in:

Code: Select all

(eq1) Ft = a + bXt+ ERRORt (eq2) Zt = c + ... + (Ft+1 - a - bXt+1) + ...