Forecasting with VEC Models
Posted: Wed Nov 11, 2009 12:29 pm
Hello.
I am new to using MODEL solving. So, any help would be greatly appreciated.
1. First, I fit a VEC model with three variables X, Y and Z (with two cointegrating equations)
- Sample size: 2006m01 to 2009m09
2. Then I use, PROC-MAKE MODEL to define and solve a model
3a. I want forecast from 2009m10 till 2015m12. I am interested in the forecast values of X and Y after fixing the variable Z at $100 (for example)
3b. I do this by entering the value 100 in Z data-points from 2009m10 till 2015m12 - IS THAT THE RIGHT WAY?
4. Then I solve it - I want multi-step forecasting; so, I guess I have to choose DYNAMIC. Should I use DETERMINISTIC/STOCHASTIC forecast?
5. If I change Z = 100, 200, 300, I obtain different forecast values for X which can be denoted as X_100, X_200, X_300
But, the forecast values for Z_100 are very different from the value Z = 100, that I entered for Z from 2009m09 till 2015m12 - Is it because the VEC model fits poorly? I am stuck.
Any tips would be helpful.
Thanks
I am new to using MODEL solving. So, any help would be greatly appreciated.
1. First, I fit a VEC model with three variables X, Y and Z (with two cointegrating equations)
- Sample size: 2006m01 to 2009m09
2. Then I use, PROC-MAKE MODEL to define and solve a model
3a. I want forecast from 2009m10 till 2015m12. I am interested in the forecast values of X and Y after fixing the variable Z at $100 (for example)
3b. I do this by entering the value 100 in Z data-points from 2009m10 till 2015m12 - IS THAT THE RIGHT WAY?
4. Then I solve it - I want multi-step forecasting; so, I guess I have to choose DYNAMIC. Should I use DETERMINISTIC/STOCHASTIC forecast?
5. If I change Z = 100, 200, 300, I obtain different forecast values for X which can be denoted as X_100, X_200, X_300
But, the forecast values for Z_100 are very different from the value Z = 100, that I entered for Z from 2009m09 till 2015m12 - Is it because the VEC model fits poorly? I am stuck.
Any tips would be helpful.
Thanks