The RATS Software Forum

Hi Tom,

I am using @MVGARCHFore.src with mv=standard to get about 10,000 one step ahead forecasts. The code works fine for the first 3000 forecasts and then gives the following error message:

## MAT2. Matrices with Dimensions 3 x 3 and 0 x 1 Involved in * Operation
Error was evaluating entry 3529
The Error Occurred At Location 2062, Line 97 of MVGARCHFORE
Called From Location 188, Line 4 of loop/block

When mv=vech or dcc the same error message occurs after 5000 or 900 forecasts respectively. I have checked and I do not think there is a problem with the data. Can you suggest a way to address this?

Also, I understand @MVGARCHFore.src cannot be used with an asymmetric model. I tried using @mvgarchtovech.src (which accepts asymmetry) and then @MVGARCHFore.src but it did not work. Are there any alternative ways to obtain the forecasts with asymmetry?

Many thanks

10,000 one-step forecasts of what? Rolling estimates? @MVGARCHFORE assumes that you are forecasting starting at the end of the estimation period. It's possible that the rolling estimates are reaching the point that they don't converge. The fact that they fail at different points for different models would seem to bear that out. You may need to check for problems with the GARCH instruction at the point where this is happening.

Hi Tom,

Is it possible to forecast the DBEKK garch using mvgarchfore ?

Thanks