Adding forgetting factor to autoregression

Hi

Im using AR() for prediction. Is there a way to add forgetting factor? I.e. the determination of the unknown model parameters should rely on the past samples with a decreasing factor, so that older samples are "forgotten".

Morten

[Morten]

I have now found the rarx(), which can estimate the parameters of the AR with forgetting factor:

[thm] = rarx(y,order,'ff',0.98);

thm is a vector with the estimates, but how do I then convert that to a discrete-time IDPOLY model, because that is what the predict() function want as input? Can I use the AR() function?