Recursive forecast of Multivariate-GARCH-in-Mean model

[chelseaisme]

Hi,
Can any one help on the programming of Multivariate-GARCH-in-Mean model. I'm using Eviews 6. At first my estimation sample is 1980m1 to 2000m1, forecast sample is 2000m2 to 2010m1. I need estimate a MGARCH-in-Mean model based on data 1980m1 to 2000m1, to forecast the stock return and bond return on 2000m2, as well as conditional stock variance, bond variance and their covariance on 2000m2. For the second time, I need estimate the model based on data 1980m1 to 2000m2(not rolling window) to forecast returns and conditional variances, covariances on 2000m3.There are 120 times such experiments in total. My data is in the attachment. The stock return series is named "spi",the bond return series is named "bdr". Thanks.
Ying

[trubador]

What you are asking is not an easy task. This code will guide you in estimating the model, while this one will give you the idea on how to do the forecasting afterwards. You have to modify and combine these codes to suit your need.

[chelseaisme]

I have made my own code. It's for MGARCH since MGARCH-in-Mean is not necessary for me. But my code cannot recursively forecast the stock and bond returns, and the forecasted variances and covariances are the same as the actual ones. Can you make some corrections on my code? Thanks.

[trubador]

You have to save the output after each loop, otherwise you will only be cross-checking the results...