Different Results

[turkhanali]

Dear Friends,

When I using FM-OLS.SRC and FM.SRC , I found the results are different. Then I checked with GAUSS program, the results from FM.SRC and Gauss code are same. I am wondering is it because the FM.SRC and Gauss code use pre-whitening before do fully modified OLS, but FM-OLS does not. Is that correct? Is there any way that we still can estimate it using FM-OLS with pre-whitening process. I have checked the FM-OLS code there is no option for pre-whitening.

In addition, Is it possible for FM.SRC do estimation without pre-whitening? How?

Thank in advance.

[TomDoan]

Any estimation that involves calculation of a long-run variance (such as fully modified OLS) will depend upon the method chosen for computing that. That doesn't make one right and one wrong; they're just different. We (at Estima) wrote fm.src to match the Hansen Gauss routine; we did not write fmols. Is there some feature in fmols that you need that's not in fm?

[turkhanali]

Hi, Tom,

Is it possible to include the structural break tests (SupF, Mean F and Lc) in the FM.SRC as additional feature of FM.src to match the Gauss code?

Or Does RATS have it already in other procedures? Please kindly advise.

Thanks.

[TomDoan]

Isn't that what Kai Carstensen's procedures do?