[R-sig-Geo] RE : CAR model with geoRglm

[Julien Beguin]

Thank you Paulo. Indeed, it seems a very VERY promising approach for a large kind of spatial models and sounds like a revolution, at least for me. 

Julien
----------------
Ph.D. Student
Laval University

________________________________________
De : Paulo Justiniano Ribeiro Jr [paulojus at c3sl.ufpr.br]
Date d'envoi : 24 mars 2010 18:26
À : Julien Beguin
Cc : r-sig-geo at stat.math.ethz.ch
Objet : Re: [R-sig-Geo] CAR model with geoRglm

Dear Julien

geoRglm fitis continuoaus spatial models and therefore CAR is not
included.
One recent and exclente alternative is the INLA package
(not in CRAN) available at www.r-inla.org
which fits a wide range of Markov random fields models

The web-site has intructions and example.
More details in the JRSS-B read paper by Rue, Chopin and Martino (2009)


Paulo Justiniano Ribeiro Jr
LEG (Laboratorio de Estatistica e Geoinformacao)
Universidade Federal do Parana
Caixa Postal 19.081
CEP 81.531-990
Curitiba, PR  -  Brasil
Tel: (+55) 41 3361 3573
Fax: (+55) 41 3361 3141
e-mail: paulojus AT  ufpr  br
http://www.leg.ufpr.br/~paulojus



On Wed, 24 Mar 2010, Julien Beguin wrote:

> Dear all,
>
> I am fitting a CAR model in WinBugs with proportional data (binomial distirbution) as response and a couple of independent variables. I have 1873 cells and Bugs does not seem to like so much (1 minute by iteration) when I estimate a spatial random effect for each cell (1st order neigboorhood).
>
> I have heard about the geoRglm package for making spatial binomial regression using MCMC estimations, but never used it. Is someone know if this method could easely handle 1873 cells (at least faster than WinBugs) or I am better to be patient and continue with Bugs.
>
> Can you give me some advice please? I would really appreciate it.
>
> Cheers,
>
> Julien Beguin
> ----------------
> Ph.D. Student
> Laval University
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>