[Biomod-commits] Explanatory Variables importance stepwise selection

[Wilfried Thuiller]

Well… I do not think you need since each algorithm is doing it for you. 

My 2-cents on that: 

1- make a careful selection prior modeling
2- run biomod2 and look at the importance of the variables.

If you really want to do a post-selection: 

3- look at which variables are consistently under-selected or have consistently a low variable importance. Remove them from the list of variables
4- re-run biomod2 with the restricted set of variables.

Hope it helps

Wilfried



Le 9 avr. 2013 à 18:31, Joaquin DL <jduquelazo at gmail.com> a écrit :

> Then, How can I do a proper variable selection, using Biomod?
> Should I used just one single model to test my relevant variables? and later apply the diferent models techniques?
> 
> Thanks
> 
> 
> 2013/4/9 Joaquin DL <jduquelazo at gmail.com>
> Dear Wilfried,
> 
> Thanks for your reply,
> As usual it is really helpful!
> 
> Joaquín
> 
> 
> 2013/4/9 Wilfried Thuiller <wilfried.thuiller at ujf-grenoble.fr>
> Dear Joaquín,
> Each algorithm has its own variable selection or optimization. For instance, both GLM and GAM are implemented with a stepwise procedure. Variables that are not important in terms of explained deviance (and resulting AIC of the models) are not kept. This is the same for the tree-based approaches.
> The variable importance procedure that biomod extracts,  which is consistent through the different algorithms, are only on the selected variables for each model. It explain why sometimes, the variable importance is 0 for some models (because the variable was not selected by the models).
> Hope it clarifies your thoughts.
> Wilfried
> 
> 
> 
> Le 9 avr. 2013 à 17:52, Joaquin DL <jduquelazo at gmail.com> a écrit :
> 
> > Dear all,
> >
> > Does Biomod any procedure to delete not relevant variables from the
> > explanatory dataset?
> > I know that there are a function "getModelsVarImport" where i can see the
> > importance of each variable in each model. But It would be nice to have a
> > function that highlight the variables to delete in each run.
> >
> > I am working with the last biomod2 version.
> >
> > Thanks
> >
> > Joaquín
> > _______________________________________________
> > Biomod-commits mailing list
> > Biomod-commits at lists.r-forge.r-project.org
> > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/biomod-commits
> 
> -----------------------------
> Dr. Wilfried Thuiller
> Laboratoire d'Ecologie Alpine, UMR CNRS 5553
> Université Joseph Fourier
> BP53, 38041 Grenoble cedex 9, France
> tel: +33 (0)4 76 51 44 97
> fax: +33 (0)4 76 51 42 79
> 
> Email: wilfried.thuiller at ujf-grenoble.fr
> Personal website: http://www.will.chez-alice.fr
> Team website: http://www-leca.ujf-grenoble.fr/equipes/emabio.htm
> 
> ERC Starting Grant TEEMBIO project: http://www.will.chez-alice.fr/Research.html
> 
> 
> 

-----------------------------
Dr. Wilfried Thuiller
Laboratoire d'Ecologie Alpine, UMR CNRS 5553
Université Joseph Fourier
BP53, 38041 Grenoble cedex 9, France
tel: +33 (0)4 76 51 44 97
fax: +33 (0)4 76 51 42 79

Email: wilfried.thuiller at ujf-grenoble.fr
Personal website: http://www.will.chez-alice.fr
Team website: http://www-leca.ujf-grenoble.fr/equipes/emabio.htm

ERC Starting Grant TEEMBIO project: http://www.will.chez-alice.fr/Research.html