Maximum lag and lead in model class

Jan 29, 2015 at 2:12 AM

Does IRIS have a maximum number of lag and lead in model class?

The maximum lag in my model is 20, but the maximum lag in xVector (with get(...,'xVector')) is 19. The same thing happens with the maximum lead.

Jan 29, 2015 at 5:00 AM
Besides the RAM on your computer there should be no constraints. In testing I am also unable to encounter any restrictions. I am able to solve models with leads and lags of 1000 on my little old laptop.
Jan 29, 2015 at 12:18 PM
There is, of course, no limit on lags/leads. This is simply the standard way how first-order difference state space equations are constructed in models with more than t-1 lags (and this has nothing to do with IRIS).

If you have a model like
w = 0.8*w(t-1) + 0.4*w(t-2) + epsilon
your state vector, i.e. xVector, is x := [w; w(t-1)], i.e. always one lag less than what appears in the equations. This is because the first-order difference matrix/vector version corresponding to the above equation is given by
A x = B*x(t-1) + D*epsilon
where A = [1,0 ; 0,1] and B = [0.8,0.4 ; 1,0]. The t-2 lag pops up automatically on the right-hand side, and hence there is no need to include it in the x vector.

Marked as answer by jaromirbenes on 1/29/2015 at 5:19 AM
Jan 29, 2015 at 5:21 PM
I should have think about the rewriting in VAR(1).

Sorry for this too simple question.