Dear Roosbeh,
Alternatively you might want to use spatial filtering, which should pretty
much accommodate any type of regression model. In spdep there's already a
function for that, running on glm, but I doubt that covers multinomial
logit. But if a stepwise (step command) works on the multinomial model (see
package mlogit) (i.e., if the command has a likelihood method), then you can
use it.
For the logistic regression, see the paper by Griffith: Griffith D A, 2004,
"A spatial filtering specification for the autologistic model" Environment
and Planning A 36(10) 1791 ? 1811
For R code on implementing the spatial filtering approach, see some examples
on my webpage: www2.dse.unibo.it/patuelli/ (look for R code).
Cheers
Roberto
********************
Roberto Patuelli, Ph.D.
Dipartimento di Scienze Economiche (Department of Economics)
Facolt? di Economia sede di Rimini (Faculty of Economics-Rimini)
Alma Mater Studiorium Universit? di Bologna (University of Bologna)
via Angher? 22
47921 Rimini
Italy
Phone: +39-0541-434276
Fax: +39-02-700419665
Email: roberto.patuelli at unibo.it
Homepage: http://www2.dse.unibo.it/patuelli/
********************
New article published in the Journal of Regional Science:
Patuelli, R., N. Schanne, D.A. Griffith and P. Nijkamp (2012). Persistence
of Regional Unemployment: Application of a Spatial Filtering Approach to
Local Labor Markets in Germany. Journal of Regional Science 52 (2): 300?23