[R-sig-Geo] spml, spfeml and tolerance level

[Ariel Ortiz-Bobea]

Hello,

I'm running various non-nested J-tests from various spatial error
fixed-effect models using the spml() function from the splm package. I'm
having trouble running some of these tests because there is collinearity
between regressors in the augmented test regression. I get the usual:

Error in solve.default(crossprod(xco), crossprod(xco, yco)) : 
  system is computationally singular: reciprocal condition number =
6.34679e-28

I'm wondering how can I decrease the tolerance so that I can override this
error. I looked inside the spml function and it calls the spfeml() function
as can be seen in R-forge:
https://r-forge.r-project.org/scm/viewvc.php/pkg/R/spfeml.R?view=markup&root=splm&pathrev=126

There I found that spfeml fixes tol.solve = 1e-10 and thus the tol.solve
parameter is not passed down by  spml() function.

I presume this is the source of singularity error I'm getting and that I
could circumvent this by setting a lower tol.solve.

Could someone suggest a way to do this, and if I'm totally wrong, offer a
clarification and some suggestions?

Thanks a lot in advance.

Ariel



-----
Ariel Ortiz-Bobea
PhD Candidate in Agricultural & Resource Economics
University of Maryland - College Park
--
View this message in context: http://r-sig-geo.2731867.n2.nabble.com/spml-spfeml-and-tolerance-level-tp7581089.html
Sent from the R-sig-geo mailing list archive at Nabble.com.