[19079] in s-news-athena
[S] S+GARCH EWMA1 forecast
daemon@ATHENA.MIT.EDU (=?iso-8859-1?q?Manos=20Venardos?=)
Wed Sep 1 05:00:02 1999
Message-Id: <19990901085734.24229.rocketmail@web601.yahoomail.com>
Date: Wed, 1 Sep 1999 09:57:34 +0100 (BST)
From: =?iso-8859-1?q?Manos=20Venardos?= <evenos@yahoo.co.uk>
To: s-news@wubios.wustl.edu
Mime-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 8bit
Hi everyone,
I am using S-PLUS 4.5 Release 2 and S+GARCH version 1.1 release 2.
I have the usual dataset
dat<-cbind(hp.s,ibm.s)
and I estimate a multivariate GARCH EWMA1 model using
the command
mod<-mgarch(dat~1,~ewma1)
I predict the covariance matrix 1000 periods ahead using the command
mod.pred<-predict(mod,1000)
When I plot the variance forecasts using for example
tsplot(mod.pred$sigma.pred[,1]^2)
I see that the variance forecast collapses to zero for both series
while the correlation forecast is constant.
Looking at the manual of S+GARCH it seems to me that the EWMA1 model is
equivalent to IGARCH(1,1) with the constant set to zero in the variance
equation.
If this is the case shouldn't the forecast of the conditional variances
be constant and equal to the last conditional variance in my sample?
i.e. Shouldn't the variance be a martingale under this specification?
This can be also found in Bollerslev "Stationarity and Persistence in
the GARCH(1,1) model".
If anyone knows anything about this I would appreciate a reply.
Thank you in advance
Manos
____________________________________________________________
Do You Yahoo!?
Get your free @yahoo.co.uk address at http://mail.yahoo.co.uk
or your free @yahoo.ie address at http://mail.yahoo.ie
-----------------------------------------------------------------------
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
