[Biomod-commits] Variable Importance Function

[Wilfried Thuiller]

Dear Joaquin,
This is a simple permutation scheme, which more or less reproduces what is done in Random Forest for example. The only difference is that RandomForest performs the permutation over the test data, whereas in biomod2 we do that on the calibration data (although we could also allow the possibility of doing that on the evaluation data in the future). 

You may take a look at this paper:
Gromping U. 2009. Variable Importance Assessment in Regression: Linear Regression versus Random Forest. American Statistician 63: 308-319.

I do not believe permutation is necessarily the best approach for each algorithm, but this is the only one that they can be applied to all algorithms in a standard and reproducible way, hence to compare the importance across the different models. 

Hope it helps

Wilfried







Le 13 mai 2013 à 11:24, Joaquin DL <jduquelazo at gmail.com> a écrit :

> Dear all,
> 
> I am really interested in the variable importance function and it
> performaces in side Biomod.
> Could you suggest any reference where i could read more about it?
> 
> 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