Student-t and Quasi-Maximum Likelihood

[tpr3396]

Hi guys,
I'd like to ask a question about error distribution when estimate GARCH model.
If I have an exchange rate return series and it's error term appears fat tails, I decide to estimate GARCH model under the assumption of Non-normality and Student-t distribution, but some literatures apply quasi-maximum likelihood (QML) to estimate the models with the same assumption. From my understanding, the QML can be computed by select the Heteroskedasticity Consistent Covariance option, but this option is only available if I choose conditional normal as the error distribution. The eviews' help also said that I should use this option if I suspect the residuals are not conditional normally distributed. Therefore, I'm confused about the Student-t and Heteroskedasticity Consistent Covariance option in eview. Could anyone help me with this? Thanks

[trubador]

In estimating ARCH models, EViews uses maximum likelihood method under the assumption that the errors are conditionally normally distributed by default. So, if I were you I would continue with the Student-t distribution...

[izempe]

Hello,

I have the same problem as you,
as far as i know and with respect to what i have red,
you can use normal distr or Semi Likelihood in the particular case in which you think that data follows t-student distribution.
So in the second case you just change the procedure that Eviews uses and thus you don't need to change the distribution

However, i am not sure, i am waiting your opinion

Best Wishes