Prev: Mamma Mia!
Next: how do I use "SelectedObject" property?
From: tzuohann on 12 Feb 2010 23:33
I've been trying to do this for a while now and am half wondering if
its a custom job.

I'm trying to do the following:

I have a 1000x2 vector of normal random variables. Call this INNOV
(for innovations)
From these, I create my VAR(2) process with Y(t) = PHI*Y(t-1) +
I*INNOV.
PHI is [0.5, 0.5; 0, 0.9] Y is a 2x1 vector. Initialize Y(0) = 0.

So, I now have my Y(t) and I know what process generated it.

Now, I want to see if a VARMA(2,2), or VARMA(2,1) model fits the data
better using BIC from AICBIC.

However, there doesn't seem to be a built in package for VARMA
estimation WITHOUT first converting into a VAR or VMA.

I tried to do it with a grey box state space estimation but the
parameters estimated were really bad when I tried to estimate a
VAR(2,0) which is exactly what the process is. The system
identification toolbox didn't do great either.

Is there some way to do this? Econometrics toolbox doesn't do it...
Wondering if this is worth coding up with some Maximum Likelihood and
posting it for corrections/updates.

Thanks
Tzuo
 | 
Pages: 1
Prev: Mamma Mia!
Next: how do I use "SelectedObject" property?