[R-sig-Geo] “stsls”

[Roger Bivand]

On Tue, 17 Apr 2012, Joseph Duncan wrote:

> Dear list
>
> I have read ?stsls? help file and Kelejian and Prucha (1998) for more 
> than one week, but still do not understand how to estimate "generalized 
> spatial two-stage least squares model" in sample data
>
> gp2a <- stsls(log(CMEDV) ~ CRIM + ZN + INDUS + CHAS + I(NOX^2) + I(RM^2) +
> AGE + log(DIS) + log(RAD) + TAX + PTRATIO + B + log(LSTAT), data=boston.c,
> nb2listw(boston.soi))
>
> Which are endogenous variables and which are exogenous variables? Which 
> variables are used to obtain 2SLS estimator?  What does the equations 
> looks like?  exactly the same as Kelejian and Prucha (1998) ?

As you can see from the help page (and the code which is the clearest 
description), the coefficient on Wy is fitted using the lagged X variables 
WX and WWX. What is endogenous or not is a matter of terminology, and may 
not actually match the variables you have. In stsls, you don't get to 
define anything, it is simply as given by Kelejian and Prucha, that is 
first step fit WY ~ [WX, WWX], then second step. You can choose only to 
use WX instead of [WX, WWX].

Hope this clarifies,

Roger


>
> Thanks
>
>
>
>
>
> Best regard
>
> 	[[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