Search All of the Math Forum:

Views expressed in these public forums are not endorsed by NCTM or The Math Forum.

Notice: We are no longer accepting new posts, but the forums will continue to be readable.

Topic: [Q] Differing Values of Loss Function in ARMAX Model
Replies: 0

 Steve Lutes Posts: 12 Registered: 12/7/04
[Q] Differing Values of Loss Function in ARMAX Model
Posted: Jun 20, 1996 11:10 AM

Dear Mathlab Newsgroup Members:

I have a question about the value of the loss function given by the
command
present(Theta-matrix) and the same loss function computed from actual
data. I'm using the ARMAX function with nn=[2 1 3 0] i.e. the AR order
is 2, the input order is 1, the noise model or MA part order =3 and the
lag is 0.

According to a document faxed to me by the Mathworks, the loss function
V is given by the formula V=e*e'/length(e). I interpet this to mean-

V = (Actual-Estimate)*(Actual-Estimate)'/number of data points

When I run the ARMAX model specified above, the present(...) command
gives me a loss function V=0.03215. If compute the loss function using
the acutal data and the estimate from idsim, I get a loss function of
V=10.9045

Can someone tell me what's happening?

Thank you

Steve Lutes
Center for Balance Disorders Rm. NA315
Dept of Oto
Baylor College of Medicine
One Baylor Plaza
Houston, TX 77030
Vox: 713-798-6336
Fax: 713-798-8658