out-of-sample forecast using rolling sample
Posted: Fri Aug 27, 2010 12:26 pm
Hello,
I have a problem using EViews 7 in estimating volatility using rolling sample. It's really important for me to figure this out since I cannot finish the empirical part of my dissertation.
My data is 15 years daily log-returns from an index (1994-2009, 3879 observations). I used the first 10 years (1994-2004, 2649 obs) to estimate the GARCH models and with the rest 5 years (2005-2009, 1230 obs) I run a rolling regression in order to estimate my volatility and evaluate my models. I downloaded the add-ins and used the advanced rolling regression choosing fixed window length, C as the variable that changes and the window size 2649 and step 1, since I want to perform 1230 out-of-sample forecasts. The output is 1230 coefficients for the period 2005-2009.
I think these coefficients are for the mean equation and what I have to do is to re-estimate the mean equation and use it to re-estimate GARCH and get the estimated volatility. However, I don't know how do perform this. My objective is to have a series of the true volatility and a series of the estimated in order to calculate the mean square error, the mean absolute error and so on.
I would be glad if you can give me an advice how to solve this problem.
Thanks
I have a problem using EViews 7 in estimating volatility using rolling sample. It's really important for me to figure this out since I cannot finish the empirical part of my dissertation.
My data is 15 years daily log-returns from an index (1994-2009, 3879 observations). I used the first 10 years (1994-2004, 2649 obs) to estimate the GARCH models and with the rest 5 years (2005-2009, 1230 obs) I run a rolling regression in order to estimate my volatility and evaluate my models. I downloaded the add-ins and used the advanced rolling regression choosing fixed window length, C as the variable that changes and the window size 2649 and step 1, since I want to perform 1230 out-of-sample forecasts. The output is 1230 coefficients for the period 2005-2009.
I think these coefficients are for the mean equation and what I have to do is to re-estimate the mean equation and use it to re-estimate GARCH and get the estimated volatility. However, I don't know how do perform this. My objective is to have a series of the true volatility and a series of the estimated in order to calculate the mean square error, the mean absolute error and so on.
I would be glad if you can give me an advice how to solve this problem.
Thanks