topic Re: GRS Test in SAS in SAS Forecasting and Econometrics

GRS Test in SAS

MasterStudent_A — Sun, 17 Oct 2021 20:18:30 GMT

Hello everyone,

I`m trying to replicate the FF-3 factor model. Unfortunately I cannot apply the GRS-test in SAS.

Anyone who could help me with the GRS-test? I didn't find any hint how to apply the GRS-test in SAS.

I´m looking forward for the help of the SAS Community.

Thank you in advance!

Re: GRS Test in SAS

PaigeMiller — Sun, 17 Oct 2021 21:41:32 GMT

My quick search for GRS test in SAS turns up nothing. Have you tried using internet search tools for this?

Re: GRS Test in SAS

MasterStudent_A — Sun, 17 Oct 2021 21:49:33 GMT

Yes, I already did.

On Google I could find some websites and tutorials guiding how to apply the GRS-test in R, Stata or Python, but unfortunately nothing regarding SAS.

Thank you!

Re: GRS Test in SAS

MasterStudent_A — Mon, 18 Oct 2021 21:58:33 GMT

GRS stands for Gibbons Ross Shanken. The GRS test is a statistical test of the hypothesis that αi = 0 ∀i.
It is a test that some linear combination of the factor portfolios is on the minimum variance boundary.
It is used by FF to test whether the expected values of all intercept estimates are zero.

Re: GRS Test in SAS

SASKiwi — Mon, 18 Oct 2021 22:53:02 GMT

I don't see that listed here in FAStats but you might this link useful.

Re: GRS Test in SAS

Ksharp — Tue, 19 Oct 2021 13:04:46 GMT

It looks like GRS Test is for the time series analysis which is under SAS/ETS .

Maybe you could post your question at Forecasting forum , some expert about time series analysis might know the answer.

Or calling @Rick_SAS @StatDave

Re: GRS Test in SAS

MasterStudent_A — Wed, 20 Oct 2021 14:28:59 GMT

It is now in the Forecasting forum, isn't it? Thank you for you help @Ksharp!

Re: GRS Test in SAS

sbxkoenk — Sat, 20 Nov 2021 20:41:25 GMT

Hello,

What SAS PROCedure are you using??

The Gibbons Ross Shanken (GRS) test is (I think) a statistical F-test for the hypothesis that all the alphas (from a set of time-series regressions) are zero.

I think that F-test is done in the procedure that you are using.

If not, it is certainly given for each individual time-series regression and you can then make your own composed / overall test.

Regards,

Koen