Hi Allin,

thanks for the hints!

I now asked Roodman to give a clear 'recipe' for his overidentification tests and whether he already knows about the differences between his method and other well established software packages.
I gonna report back if there is some news.

In terms of the Baltagi data set it's as documented on page 143 in the gretl user-guide.
Restricting the sample length makes the point estimates of one-step and two-step very close to each other.

I now know how to perform the Difference-in Sargan test.
When the accessors gonna be available then it could be coded as well.
So long it's with printing the restricted and unrestricted model and saving the test statistics manually and perform the calculations.


01.07.2013 04:10, Allin Cottrell:
On Sun, 30 Jun 2013, Pindar wrote:

I'm still struggling with the dpanel methodology and the comparison of 
results to e.g. Stata.
First, the Sargan test statistics reported by GRETL are equivalent to the 
ones of Arellano and Bond (1991) Sargan tests.
Yes; and in most cases they are identical with those produced by 
Ox/DPD. In gretl we use the formula given in the DPD manual to 
compute the Sargan test -- maybe this should be given in the User's 

The assertion that the Sargan test of GRETL is the Hansen test in xtabond 
seems not to be true for *xtabond2*.
In some cases the assertion holds true, maybe not in others.

GRETL values are always closer to the Sargan tests of Roodman reported in 
Roodman (2006). What is the Hansen test then?
Ask Roodman, or another Stata guru. I don't know. It's not 
adequately documented in the xtabond2 PDF file.

In Baltagi (2005) I found a xtabond output. Here the results for GMM-Diff 
one-step estimates are the same as of gretl and the Sargan test fits too 
(note, here is only a Sargan test is reported in the output).
Strange in this comparison: In GRETL the two-step estimators are far away 
from the one-step coefficients and completely different to the ones reported 
in Baltagi (p. 157).
This is a case where the "A" matrix is singular and so -- as 
explained in the Gretl User's Guide -- all bets are off. Gretl and 
Ox/DPD do the same thing (generalized inverse, Moore-Penrose). Stata 
apparently does something else, we don't know what.

Another questions is how to perform the 
Difference-in-Sargan/Hansen tests in GRETL (as reported in 
At this point you'd have to code that yourself.

Allin Cottrell
Gretl-users mailing list