Endogenous explanatory variables in sphet
On Mon, 6 May 2013, Feichtinger, Paul wrote:
Dear list members, I was wondering if someone could help me out. Estimating a general spatial model (gstslshet) our model is likely to suffer from endogeneity not only introduced by the spatial lag but also from other explanatory variables. Is there any possibility to explicitly account for this endogeneity (by using specific instruments)? Is it covered/dealt with in some kind? Obviously I do not understand equation 4 in Piras (2010). I
I suggest that you install the development version of sphet from R-forge:
install.packages("sphet", repos="http://R-Forge.R-project.org")
and look for spreg(), which is a wrapper function for the model-fitting
functions. You should look for the endog= and instruments= arguments,
where:
spreg(formula = CRIME ~ HOVAL, data = columbus, listw = listw,
endog = "INC", instruments = "DISCBD")
is an example of usage.
Hope this helps,
Roger
read Anselin, Amaral and Arribas-Bel (2012) where they compare different GMM implementations among others also with an endogenous explanatory variable, but cannot reproduce the results in R sphet (problem is knitr, sweave,... I never use that), which is the only spatial econometrics software I use as I am a beginner. I would be really glad I someone could explain me this issue. Best regards, Paul --- Technische Universit?t M?nchen - TUM Paul Feichtinger Lehrstuhl f?r Volkswirtschaftslehre, Umwelt?konomie und Agrarpolitik Alte Akademie 14 85354 Freising Tel: +49.8161.71.3574 Fax: +49.8161.71.3408 e-mail: Paul.Feichtinger at tum.de [[alternative HTML version deleted]]
Roger Bivand Department of Economics, NHH Norwegian School of Economics, Helleveien 30, N-5045 Bergen, Norway. voice: +47 55 95 93 55; fax +47 55 95 95 43 e-mail: Roger.Bivand at nhh.no