Page 1 of 1

Alternating GMM estimators

Posted: Mon Aug 31, 2009 1:31 pm
by arturobandini
Hi everybody,

I'm running a self-written procedure that has among others two equations that I estimate with GMM. I would like to alternate between the two-step, iterative and continuously-updating GMM estimators, as well as with different bandwidth specifications of the Newey-West estimator in order to assess possible discrepancies in the estimations.

For instance, to estimate the two-step GMM with fix bandidth=12 according to the instructions of the EViews Command Reference, I write:

eq.gmm(c,b=12) y c x1 x2 x3

However, it doesn't change anything to the results the output still says "Simultaneous weighting matrix & coefficient iteration" and "Kernel: Bartlett, Bandwidth: Fixed (3), No prewhitening".

Does anybody see what could possibly be wrong?
Thx so much in advance!
Art.

Re: Alternating GMM estimators

Posted: Mon Aug 31, 2009 3:25 pm
by EViews Gareth
Nothing wrong with your syntax that I can see. You should post your program/workfile.

Re: Alternating GMM estimators

Posted: Mon Aug 31, 2009 3:41 pm
by arturobandini
Thx for the quick reply!
So here's the programm and the workfile. Just ask if there's anything unclear! The two gmm equations I mentionned are about halfway down in the programm...

Best, Art.

Re: Alternating GMM estimators

Posted: Mon Aug 31, 2009 4:21 pm
by EViews Gareth
Nowhere in your program do I see a GMM command that sets the bandwidth.

In fact I can't find any GMM commands that have any options at all....

Re: Alternating GMM estimators

Posted: Mon Aug 31, 2009 4:32 pm
by arturobandini
Thx for the reply. You're correct.

Here's the right version with the two gmm equations with option for two-step estimator and bandwidth=12 that do not seem to work...

Thx a lot, Art.

Re: Alternating GMM estimators

Posted: Mon Aug 31, 2009 4:59 pm
by EViews Gareth
Every time you estimate an equation via GMM with options, you then immediately re-estimate it by GMM with no options.

At a guess where you have:

Code: Select all

equation acils1.gmm libor3m c cpi(+k) gap fxgap @ cpi(-1 to -6, -9, -12) gap(-1 to -6, -9, -12) fxgap(-1 to -6, -9, -12) freeze(mode=overwrite, tableci1) acils1.gmm
You probably want:

Code: Select all

equation acils1 freeze(mode=overwrite, tableci1) acils1.gmm libor3m c cpi(+k) gap fxgap @ cpi(-1 to -6, -9, -12) gap(-1 to -6, -9, -12) fxgap(-1 to -6, -9, -12)
or something similar

Re: Alternating GMM estimators

Posted: Mon Aug 31, 2009 5:18 pm
by arturobandini
Thx for the reply again! This is indeed where the problem was... I thought that it was not necessary to restate your options when you just want to store the estimates into the table. Good to know!

By the way, a suggestion of mine regarding the Commands Reference might be to link clearly the different options to the different gmm estimators as they are exposed in all textbooks.

My guess is: "o"= two-step gmm (hansen 1982), "s"= iterated gmm (Ferson and Foersterb, 1994), "i"=continously-updating gmm. But then I don't know what I should associate with "c"...

Another aspect to clarify would be wether those different options change the first-step estimate of the weighting matrix. The Eviews Guide II says it is the consistent TSLS estimator but I was not sure if it was indeed the case for all these options (another option might be the identity matrix...).

Anyway, thank you very much!
Art.

Re: Alternating GMM estimators

Posted: Mon Aug 31, 2009 5:40 pm
by EViews Gareth