Page 1 of 1

ECM Forecast

Posted: Thu Nov 18, 2010 1:09 pm
by kosovard
Hi! I apologize for my bad english...

I am trying with eviews 7 and with a ECM method to "forecast" a variable.
Here is the model:
  • LT relationship: log(y)=a+b.log(x) +Ut
    ST relationship: dlog(y)= c + e.dlog(y-1) + f.(Ût-1) + EPSILONt
    putting LTequation into ST equation I find log(y)=c+ (a+b.log(x-1)).(d+1) -d(a+b.log(x-2)) +e.(Ût-1)
    Hyp:Ut and Epsilont are noises (I hope there are no mistakes in the last equation)

Code: Select all

equation.ls log(y) c log(x) genr Ut=resid equation.ls dlog(y) c dlog(y(-1)) Ut(-1)
I have no problem to estimate the first and the second regression but what I want to do is to forecast log(y) thanks to these two equations and i have not found the way to do it with eviews. The way I want to forecast is defined in the last equation. Maybe there is an other way to retake the coefficient from the output window (I can't do it manually because I have a lot of this type of ECM regression and a lot of forecat to do). Is there an easy code to take the coefficient and re-integrate them into an equation?

Please note that I do not want to forecast short term and long term relationship

So, I just need to generate a second version of log(y ) for all the values of the serie x thanks to the regressions I have made and the coefficients i have found.


I hope I've been clear, if not just tell me, I'll try to explain better.

Thank you for your help,

Kos

Re: ECM Forecast

Posted: Fri Nov 19, 2010 4:33 am
by kosovard
Problem solved, note that I have re-specified my model which was wrong.

Tanks