MV-GARCH-M with square roots

[TomDoan]

This is an example of a three variable multivariate GARCH-M model where the "M" effect are loadings on the square roots of the variances. Because the mean function is non-linear in the variances, it can't be done using GARCH with variances in the mean model. Instead, you have to use MAXIMIZE. The mean function is set in the following line:

Code: Select all

frml residv = yv-b-ar*yv{1}-gm*%sqrt(%xdiag(h))

(This also has a VAR(1) term). Other types of "M" effects can be handled in a similar fashion by changing how the "H" (which is a symmetric nxn matrix) enters into this.

garchmv-msqrt.rpf

(3.64 KiB) Downloaded 1445 times

The data set is from the standard GARCHMV example.