How to get F-stat in GARCH/ARCH estimation?

[debasishmaitra]

Please somebody help me in getting the F-stat of GARCH estimation in Eviews-7?
Its not coming in output table..

[EViews Gareth]

We dropped it from the output, since we felt it wasn't being calculated properly.

EViews 6 would compute the F-statistic by comparing the estimated equation against a constant only (i.e. no GARCH/ARCH terms) model, when in fact what you really want to do is compare it to the no-regressors (other than constant) model which has GARCH/ARCH terms. Since this would require re-estimating a model, it was decided to simply drop the F-stat.

[fmgoto]

Gareth, I have used the code below for calculating the probability of F. Although it is a good approximation, it is quite often different from the F calculated by Eviews.

Could you provide the formula Eviews uses for calculating it? Which is more accurate?

Thank You, Fabio

Code: Select all

scalar f_p_value =  1 - @cfdist(eq01.@f, eq01.@ncoef-1, eq01.@regobs-eq01.@ncoef)


[EViews Gareth]

That's exactly how it is calculated. Could you provide an example of it not matching?

[fmgoto]

I have been running this code for a while and some times I get two different values. But I dont get an example right nwo. I ll try to remember posting it next time I get it. Probably my mistake. Thanks.

[debasishmaitra]

Many Many Thanks Gareth and Fmgoto for your prompt reply..
Thanks a ton..\
But Gareth and Fmgoto can you tell me how to put this code to get F-stat? I mean where to write it..?? If you can give a protocol to do this...

[Rosa.hh]

Code: Select all

scalar f_p_value =  1 - @cfdist(eq01.@f, eq01.@ncoef-1, eq01.@regobs-eq01.@ncoef)


Hi.
I'm trying to calculate the F-stat in an EGARCH estimation using the code above however it is not working for me when using the specification below.
Any idea of what should I do to calculate the F-stat using an EGARCH model? I enclose my database

Thanks a lot,

equation eq01.arch(1,1,egarch,m=1000,h) dlner c dlner(-1 to -1) i_diff(0 to 0) BOG_APR BOG_NET BOG_DEP mon tue wed thu @ BOG_APR BOG_NET BOG_DEP mon tue wed thu

[EViews Gareth]

As mentioned above, EViews 7 does not calculate the F stat for GARCH/ARCH models.

[Rosa.hh]

Thanks for the quick response Gareth. Is there any alternative to applying a joint significance test in EGARCH estimations? Is it possible to calculate it from scratch using the F-test formula [ (RSS1-RSS2/P2-P1)/RSS2/N-P2] ?