The RATS Software Forum

hi, everyone;
do you any suggestion about how to capture data with rolling window in GARCH model? is it similar with rolling regression?

Thanks for any help

Yes. You'll just want to save the results of the %beta vector similar to what you would do with a rolling regression.

Here's an example. This uses the HESSIAN and INIT options to feed the previous estimates in (you just have to handle the first estimate differently). That tends make the estimation run faster, cutting the number of iterations roughly in half.

You need to be very careful not to make your window too narrow. If you get a quiet period where there is no obvious GARCH effect, followed by some outliers, the estimates can get very unstable. The person who originally submitted this was using a width of 250, which doesn't work well once you get around observation 400.