The RATS Software Forum

Is there any way to estimate a DCC GARCH model with different sets of regressors in the variance equations? The code for my attempt to do so is in the attached file. It's a modification of the example file "garchmvdcc2.prg."

The problem comes at the end, when the "garch" command is used for one iteration to get the grand covariance matrix. The garch command requires the same set of regressors for all variance equations in multivariate models.

I'm estimating a bivariate system, and I would like to test cross-equation restrictions, too.

There's no way to do that second stage with the full model, at least not with the GARCH instruction. You can, however, estimate the univariate models and the DCC model given the standardized residuals from those.