Forum: open-discussion

RE: Multivariate GARCH (1, 1) [ Reply ] By: Arne Henningsen on 2014-09-13 07:43	[forum:41450]
I do not know "bollerslev and wooldridge corrected SEs." Is it specific to GARCH models? You are welcome to implement this in maxLik.

RE: Multivariate GARCH (1, 1) [ Reply ] By: saikat sarkar on 2014-09-13 07:13	[forum:41449]
Hi, They mean that bollerslev and wooldridge covariance matrix. Is there any way we can estimate the bollerslev and wooldridge corrected SEs in maxLik for the Bi-variate GARCH(1, 1) model. Thanking you in Advance

RE: Multivariate GARCH (1, 1) [ Reply ]
By: Arne Henningsen on 2014-09-13 06:56

[forum:41447]

Dear Saikat

I do not know what the GAUSS documentation means with "quasi maximum likelihood covariance matrix." You can use argument "finalHessian" to choose between 2 methods to calculate the Hessian matrix that is used to calculate the covariance matrix. The calculation of the Hessian (and thus the covariance matrix) (and the estimation if you use a gradient-based optimisation method) becomes more precise when you provide arguments "grad" (and "hess" if you do not use the BHHH approximation).

Best regards,
Arne

Multivariate GARCH (1, 1) [ Reply ]
By: saikat sarkar on 2014-09-13 06:44

[forum:41445]

HI,

I have etimated Bivariate GARCH(1, 1) with maxLik function. The estimated parameters are quite same with the Gauss estimation, but standard errors are totally different from the GUASS estimate. GUASS program said that the SEs are based on quasi maximum likelihood covariance matrix.

Can you advice me how to get quasi maximum likelihood covariance SEs for each parameter estimate from maxLik function.

Any suggestion please.

Thanking you