[Openspace] LM Tests and Model Choice

[Donald Lacombe]

Dear OpenSpace Community:

	I recently read section 26.6.3 in the GeoDa workbook on the various
LM test statistics and had a few questions regarding model choice:

1) I am aware that there are identification issues when using the general
spatial model under maximum likelihood but have seen conflicting reports as
to what the real issues are, i.e. using the same versus different weight
matrices for the lag and error term, whether or not the model is identified
if there is more than one exogenous variable, etc. Can someone point to a
definitive reference as to what the salient issues are?

2) What are the pros and cons of using a GMM type general spatial model
along the lines of the Kelejian-Prucha GMM model in their 1998 paper "A
Generalized Spatial Two-Stage Least Squares Procedure for Estimating a
Spatial Autoregressive Model with Autoregressive Disturbances"? The GeoDa
workbook doesn't mention this route and basically stops at the standard
spatial lag and error models and I was wondering why the general model
doesn't get a mention.

	I want to thank users in advance for any help that can be provided.
Apologies if this has been asked before--a search option for he archives
would be a nice feature.

Donald Lacombe

*************************************
Donald J. Lacombe
Assistant Professor of Economics
Ohio Univeristy
Athens, OH 45701
Voice: (740) 593-0145
Fax: (740) 593-0181
email: lacombe at ohio.edu
*************************************