We want to compute a pseudo R-squared for a model whose parameter
estimation was based on maximum likelihood (function likfit, package geoR).
I tried to compute the R2 proposed by Maddala (1983) which compare the
maximized likelihood for the model without any predictor and the
maximized likelihood for the model with all predictors.
I got a really low value (0,01%). Did I miss something? Are there other
R-squared which are more appropriate than the R2 of Maddala? :
>sp05_l01 <- likfit(cnm05g, *ini.cov.pars=vm05*, lik.method =
"REML",
trend = trend.spatial(~ logIKA04 + nbLitre0 +
nbLitre1, cnm05g))#complete model
>sp05_nt <- update(sp05_01, trend = trend.spatial("cte",
cnm05g))
#null model
>1-(sp07_nt$loglik/sp07_l0$loglik)^(2/590)#Maddala (1983)
[1] 0.0001510074
Thanks,
Regards,
Marion
--
Marion Jacquot
Laboratoire de Chrono-environnement
UMR UFC/CNRS 6249 USC INRA
Université de Franche-Comté
Place Leclerc
F-25030 Besançon cedex
FRANCE
Tel. : +33 (0)381 665 829
Fax : +33 (0)381 665 797
http://chrono-environnement.univ-fcomte.fr/
[[alternative HTML version deleted]]
This is basically a statistics question, despite the fact that you refer to an R implementation. You should therefore probably post on a statistics list, like stats.stackexchange.com, especially the part about asking for alternative R^2. -- Bert On Wed, Nov 7, 2012 at 9:01 AM, Marion Jacquot <marion.jacquot at univ-fcomte.fr> wrote:> We want to compute a pseudo R-squared for a model whose parameter > estimation was based on maximum likelihood (function likfit, package geoR). > I tried to compute the R2 proposed by Maddala (1983) which compare the > maximized likelihood for the model without any predictor and the > maximized likelihood for the model with all predictors. > I got a really low value (0,01%). Did I miss something? Are there other > R-squared which are more appropriate than the R2 of Maddala? : > > >sp05_l01 <- likfit(cnm05g, *ini.cov.pars=vm05*, lik.method = "REML", > trend = trend.spatial(~ logIKA04 + nbLitre0 + > nbLitre1, cnm05g))#complete model > >sp05_nt <- update(sp05_01, trend = trend.spatial("cte", cnm05g)) > #null model > > >1-(sp07_nt$loglik/sp07_l0$loglik)^(2/590)#Maddala (1983) > [1] 0.0001510074 > > Thanks, > Regards, > Marion > > -- > Marion Jacquot > Laboratoire de Chrono-environnement > UMR UFC/CNRS 6249 USC INRA > Universit? de Franche-Comt? > Place Leclerc > F-25030 Besan?on cedex > FRANCE > Tel. : +33 (0)381 665 829 > Fax : +33 (0)381 665 797 > http://chrono-environnement.univ-fcomte.fr/ > > > > [[alternative HTML version deleted]] > > > ______________________________________________ > R-help at r-project.org mailing list > https://stat.ethz.ch/mailman/listinfo/r-help > PLEASE do read the posting guide http://www.R-project.org/posting-guide.html > and provide commented, minimal, self-contained, reproducible code. >-- Bert Gunter Genentech Nonclinical Biostatistics Internal Contact Info: Phone: 467-7374 Website: http://pharmadevelopment.roche.com/index/pdb/pdb-functional-groups/pdb-biostatistics/pdb-ncb-home.htm
... and you might also wish to try the R-sig-geo list, for obvious reasons. -- Bert On Wed, Nov 7, 2012 at 9:01 AM, Marion Jacquot <marion.jacquot at univ-fcomte.fr> wrote:> We want to compute a pseudo R-squared for a model whose parameter > estimation was based on maximum likelihood (function likfit, package geoR). > I tried to compute the R2 proposed by Maddala (1983) which compare the > maximized likelihood for the model without any predictor and the > maximized likelihood for the model with all predictors. > I got a really low value (0,01%). Did I miss something? Are there other > R-squared which are more appropriate than the R2 of Maddala? : > > >sp05_l01 <- likfit(cnm05g, *ini.cov.pars=vm05*, lik.method = "REML", > trend = trend.spatial(~ logIKA04 + nbLitre0 + > nbLitre1, cnm05g))#complete model > >sp05_nt <- update(sp05_01, trend = trend.spatial("cte", cnm05g)) > #null model > > >1-(sp07_nt$loglik/sp07_l0$loglik)^(2/590)#Maddala (1983) > [1] 0.0001510074 > > Thanks, > Regards, > Marion > > -- > Marion Jacquot > Laboratoire de Chrono-environnement > UMR UFC/CNRS 6249 USC INRA > Universit? de Franche-Comt? > Place Leclerc > F-25030 Besan?on cedex > FRANCE > Tel. : +33 (0)381 665 829 > Fax : +33 (0)381 665 797 > http://chrono-environnement.univ-fcomte.fr/ > > > > [[alternative HTML version deleted]] > > > ______________________________________________ > R-help at r-project.org mailing list > https://stat.ethz.ch/mailman/listinfo/r-help > PLEASE do read the posting guide http://www.R-project.org/posting-guide.html > and provide commented, minimal, self-contained, reproducible code. >-- Bert Gunter Genentech Nonclinical Biostatistics Internal Contact Info: Phone: 467-7374 Website: http://pharmadevelopment.roche.com/index/pdb/pdb-functional-groups/pdb-biostatistics/pdb-ncb-home.htm