testing GARCH(1,1)

richardd · April 14th, 2004, 8:30 am

Hi.I'm writing a GARCH(1,1) engine using Levenberg-Marquardt parameter estimation (as described in Flattery et al) for volatility updates/forcecasts (as described in Hull) and I'm looking for a standard or accepted or just reasonable method of testing it. any advicewould be greatly appreciated.Richard.

SPAAGG · April 14th, 2004, 9:59 am

Just simulte a GARCH(1,1) and then, with your time series, fit your GARCH. You should find the same parameters and have gaussian residuals, iidhope it helpsDavid

richardd · April 14th, 2004, 1:50 pm

you make it sound so easy forgive my ignorance: I'm not clear what you mean when you say 'simulate a GARCH(1,1)'. are you suggesting some kind of monte carlo simul? if so, if there's a handy reference you could point me at I'd be very grateful. as it is I've been trying to track down "Grappling with GARCH", Engle & Mezrich, but I can only find references to a september edition of RISK and have had no luck in finding a RISK repository where I might get hold of it.thanks in advance,Richard.

SPAAGG · April 14th, 2004, 2:19 pm

try to download R, from R-projectI am sure you'll find free package to simulate a GARCH(1,1) process. good luck

quantman · April 15th, 2004, 10:53 am

Hi richardd,email me your sample data and your results, I'll compare it to the results I'd get on my GARCH engine (which can also handle EGARCH and some others kind) on the same sample. I know that my engine is pretty robust as I backtested it against GARCH toolboxes (Eviews and Matlab).Tell me if it suits you.Regards.

NoDoubts · April 17th, 2004, 10:00 pm

everything is in here http://www.spatial-econometrics.com/