Re: [S] A Simple Question about Correaltion

PT Quants Group Dir (ptquants@ptqsrv1.etsd.ml.com)
Wed, 26 Aug 1998 11:45:22 -0400


A simple exponential weighting like JP Morgan Riskmetrics should work fine;
however, GARCH may be more satisfactory way to go, which means you should buy
the Garch module (Splus too slow otherwise and writing them is pain). Please
look at JP Morgan website for the info about calculating VaR. It's good
heteroscedasticity measure, but won't do much for non-linearity.

S.Y.

> This is not a purely S-plus question. I am trying to measure the association
> between two series. Historical correlation (as rescaled covariance) numbers
> are not satisfactory. Any idea how I can get numbers that are more stable
(?),
> uses the probable existence of non-linearity & heteroscedasticity in data ?
> I don't want to use anything like multivariate-GARCH at this point.
>
> Thanks.
>
> Mustaque Ahmed
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news