OLS regression with GARCH residuals

[potetbonde]

Hello.

I'm modelling the risk premium in the Nordic electricity market using weekly contracts. I have chosen to do an OLS regression. In my literature study, I read an article which did a similar regression. Although, due to extreme spikes the model explains very little of the variance. They developed two models; 1) A regular OLS model with robuste std.errors and 2) A OLS with GARCH residuals (at least that's how I interpreted it), robuste std error here as well. . The coefficients from the two regression were very similar, but the t-stats are different making other variables significant.

Could someone please explain to me how they are doing it? The paper said they used GRETL, but I was hoping Eviews could do it as well, maybe in combination with Excel.


I appreciate all help.

[nishantvats12]

Could you post the paper for us to get a better clarity?

[trubador]

Paper replication is not a simple task. You should first make sure that:
1) you use the exact same data
2) you build the exact same model
3) you choose the exact same specifications regarding the software

Even if you manage these properly, there can still be differences among softwares in terms of the optimization algorithm (e.g. method, tolerance, etc.) used in the estimation or the handling of such models (e.g. initialization).