Thank you Sven

As for the weighting matrix, I would like to access the two-step version of the AN matrix in formula (21.5) from the Guide.

in the formula of AN after (21.5) the matrix H* is considered for computation, so that, as far as I understand, this is the matrix used to obtain first step results.

The two-step version would replace the matrix H* with cross-products of first-step residuals (residuals estimated on the basis of first-step coefficients).

I hope this clarifies what I was referring to.

However,as the matrix of instruments is not available, I would need to use a different approach.

Kind regards

Laura

Date: Sat, 24 Nov 2018 14:03:46 +0100
From: Sven Schreiber <svetosch@gmx.net>
To: gretl-users@lists.wfu.edu
Subject: Re: [Gretl-users] Saved results from dpanel
Message-ID: <58f8ce83-c5ec-a7ce-2642-9dfbe58adecc@gmx.net>
Content-Type: text/plain; charset="windows-1252"; Format="flowed"

Am 24.11.18 um 10:43 schrieb Laura Magazzini:
>
> Hi!
>
> I am running a system GMM estimator for dynamic panel data models.
>
> I need to perform some computations after the second step estimation
> than involve the matrix of instruments, the matrix of regressors (both
> differenced and level), and the weighting matrix used in computation
> of the system GMM (obtained from first step estimates).
>
> Is it possible to recover these information after the dpanel command?
>

Hi, that's a very good question, and currently it seems to me that the
answer is no.

What you can get after the dpanel estimation is the list of regressors
in $xlist. However, I just checked it and it doesn't include the lagged
endogenous variables; maybe it should.

Also, while you can get the list of instruments after a regular 'tsls'
estimate via $model.instlist (a little hidden as well, maybe a direct
$instlist accessor should be added), for dpanel there is no such thing
in the $model bundle AFAICS.

So I guess you pointed out some gaps in gretl that should be filled.


About the weighting matrix: Maybe you could describe in the notation of
ch.21 of the guide which matrix exactly you are referring to.


> Would also be possible to compute the Ahn & Schmidt non linear panel
> data estimator in gretl? Should I program it by myself using gmm?
>

I am almost certain that setting up a gmm block for that estimator is
feasible in hansl, but I'm not familiar with it, so I cannot provide any
further advice there.

cheers,

sven