Page 1 of 1

Forecasting Out-of-sample data

Posted: Wed Jun 13, 2018 10:19 am
by Amdall

I am preparing thesis about the optimal hedge ratios using OLS, VECM and CCC-GARCH, BEKK-GARCH models. Let's say that my in-sample period is from 01/01/1995 to 01/01/2018 and out of sample is from 02/01/2018 to 15/05/2018. I obtained the in-sample optimal hedge ratios (static and time-varying) and I would like to obtain out-of-sample ratios as well. Can you tell me if am I correct?

1. For static out-of-sample ratios (OLS, VECM) I just use out-of-sample data and estimate models in the same way like in sample data?
Or i make a static forecast but i do not know how to obtain slope coefficient for OLS and variances and covariances for VECM using forecasted data...

2. For dynamic out-of-sample ratios (CCC/BEKK-GARCH) I have to obtain the next day conditional covariances and conditional variances somehow using out of sample data? Can u guide me a bit how to do that, because I have the same problem like one previous person on this forum and forecast (making and solving model) give me just forecasted Spot and futures returns of indices.

For this kind of forecast I am stucked with the Trubador's explanation of forecasting: viewtopic.php?t=871
It is also worth to add that I obtain the systems for GARCH models from VECM outcome.

I hope I am close to solve this issue. Thank you for any suggestions! (I have access to EVIEWS 7 and the most recent student version)