Dear Hao-pang,
it is impossible to really tell the problem without a reproducible
example. Just guessing: this looks like you have too many regressors.
In GMM, lags of variables are used as instruments, so you might have
more regressors than observations. Try reducing the 'lag' argument
(which, by default, uses all lags available).
Of course, the first observation to make would be that a GMM approach is
suited to "short", "large" panels (meaning *rather* large):
but you know
this...
HTH,
Giovanni
------------ Original message ------------------
Message: 58
Date: Sat, 31 Jul 2010 00:44:57 +0800 (CST)
From: "97258031" <97258031 at nccu.edu.tw>
To: "r-help" <r-help at r-project.org>
Subject: [R] problems about dynamic GMM
Message-ID: <1280508297.13984.97258031 at nccu.edu.tw>
Content-Type: text/plain; charset=big5
Dear all,
I am a new user of R, and I would like to use R to estimate dynamaic GMM
of Arellano and Bond (1991).
The package I used is "plm" and its code "pgmm."
However, the regression cannot run and it showed an error message:
"Error in solve.default(Reduce("+", A2)) :
system is computationally singular: reciprocal condition number
2.27327e-27."
Could I please ask, how could I deal with this problems, or are there
any other package which can help me estimate dynamic GMM?
Many thanks for help.
--
Hao-pang
----------- End original message -----------------
Giovanni Millo
Research Dept.,
Assicurazioni Generali SpA
Via Machiavelli 4,
34132 Trieste (Italy)
tel. +39 040 671184
fax +39 040 671160
