I’ve made a model from a series of (earlier estimated) equations with some common parameter names and common rhs variables:
X1= …..
X2=……
X3=……
X4=……
I then construct an auxiliary variable from that system by inserting the text
C1 = (rhs of X1 equation)/ (rhs of X2 equation)
in the equation box. Then just to check matters I created a second (seemingly identical variable):
C2 = X1/X2
When I dynamically simulate the model (with bootstrap and parameter uncertainty options allowed), I find to my surprise that C1 doesn’t equal C2. This makes me confused as to the best way to construct auxiliary variables in the model simulation option.
Auxiliary vbles in stochastic simulation environment
Moderators: EViews Gareth, EViews Moderator
-
- Fe ddaethom, fe welon, fe amcangyfrifon
- Posts: 13307
- Joined: Tue Sep 16, 2008 5:38 pm
Re: Auxiliary vbles in stochastic simulation environment
For a non-stochastic solve, they will be identical.
For a stochastic solve, you've imposed different restrictions on the randomness on C1 and C2. In C1 there is no coefficient uncertainty, since EViews doesn't know it is coming from an estimated equation. Also, remember, you will be getting different residual draws between the two calculations too, so even without coefficient uncertainty, they will not match exactly.
For a stochastic solve, you've imposed different restrictions on the randomness on C1 and C2. In C1 there is no coefficient uncertainty, since EViews doesn't know it is coming from an estimated equation. Also, remember, you will be getting different residual draws between the two calculations too, so even without coefficient uncertainty, they will not match exactly.
Follow us on Twitter @IHSEViews
Re: Auxiliary vbles in stochastic simulation environment
Thanks, that's really helpful.
From what you say, it sounds like the second method, C2 = X1/X2, is to be preferred, since you preserve the parameter uncertainty inherent in the calculation of the fitted values of X1 and X2.
Would that be the best (or most stochastically representative) practise?
From what you say, it sounds like the second method, C2 = X1/X2, is to be preferred, since you preserve the parameter uncertainty inherent in the calculation of the fitted values of X1 and X2.
Would that be the best (or most stochastically representative) practise?
-
- Fe ddaethom, fe welon, fe amcangyfrifon
- Posts: 13307
- Joined: Tue Sep 16, 2008 5:38 pm
Who is online
Users browsing this forum: No registered users and 14 guests