Obtaining loss functions of forecated volatility vs actuals
Posted: Mon Aug 08, 2022 5:57 am
Hi there,
So I am comparing volatility forecasts produced by a range of models, and want to evaluate them against a proxy for actuals (which I'm using 5-min realized variance). In obtaining my forecasted values, I estimate an ARCH/GARCH model for my mean equation for an in-sample period, then forecast the out-of-sample period and save the resultant GARCH series. However, the loss functions such as MAE,RMSE,MAPE for this forecast are for the mean equation I believe. As I have my 'actuals' saved as another series, how do I generate a forecast purely for volatility and compare it to these actuals as to obtain the correct MAE,RMSE and MAPE?
So I am comparing volatility forecasts produced by a range of models, and want to evaluate them against a proxy for actuals (which I'm using 5-min realized variance). In obtaining my forecasted values, I estimate an ARCH/GARCH model for my mean equation for an in-sample period, then forecast the out-of-sample period and save the resultant GARCH series. However, the loss functions such as MAE,RMSE,MAPE for this forecast are for the mean equation I believe. As I have my 'actuals' saved as another series, how do I generate a forecast purely for volatility and compare it to these actuals as to obtain the correct MAE,RMSE and MAPE?