MSE from GARCH forecast

Hi Shamiri,

The package will be updated during the next build cycle which means in 
about 24 hours.
The package calculates mse of variance forecast as:
mse = mean[('realized' - forecast).('realized' - forecast)]. Given a 
dataset of 1000 obs for example, and selecting to test mse on the last 
100, means that we treat those 100 as the realized observations. For 
conditional variance we usually take the actual data for those 100 obs, 
and filter (based on information from the previous 900 obs) to obtain 
residuals which you square to get 'realized' variance which you can then 
compare with the garch forecast variance.

HTH,
Alexios

Hi Alexios,
First I would like to thank you for such a package (rgarch). I downloaded the package from R-Froge, but it seems it is not updated as u said. However, since you are the right person to ask, may you help to compute the MSE for OUT-OF-SAMPLE forecasts? I am familiar with your package and how to obtain MSE for in-sample forecast. What I meant about the length of sigma2.hat, is that the length of estimated variance (model output) has the same length as my data (model input). However, I my have the wrong formula to obtain MSE.
I appreciate your help and consideration you may give me.
Regards,
Shamiri

_______________________________________________
R-SIG-Finance at stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only.
-- If you want to post, subscribe first.