[R] Package for: Instrumental variables with spatial error and lags for cross-sectional data?

cpblpublic+rhelp at gmail.com cpblpublic+rhelp at gmail.com
Mon Feb 15 16:19:11 CET 2016

I have a spatial cross-sectional dataset. I think I have an instrument for 
my dependent variable, but there's reason to suspect I need a spatial lag 
with spatial error model. The dataset has about 100,000 observations.

I am looking for a package which will estimate my model in R (or Stata or 
Python, if that will be a lot easier).

Here's where I'm stuck in R:

Package spdep has some spatial error /spatial lag models (sarar and 
gstsls), but no I.V.

Package splm has a g2sls function, which seems to offer nearly what I 
want. But this package is for spatial panels ... but not cross-sections.

Package sphet makes use of IV, but not the kind where you have your own 
instrument. Instead, it's simply using spatial lags of X as instruments 
for a spatial lag model. This seems also what this blog is about.

There's a s2sls package, i.e. spatial 2sls, version 0.1 in Jan 2016. It 
looks too new and undocumented to try.

Surely if there's a nice package for panel data IV, there's one also for 


More information about the R-help mailing list