Random forest graph interpretation

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

Random forest graph interpretation

Paula Eveline d'Anunciação
Hello, I did random forest analysis and I obtained two graphs. In the first one (left hand side) we can see the measure of importance of each predictor and on the second one (right hand side) is the proportion of variance explained by each predictor in random forest. 
I am not sure about the interpretation of some predictors. For instance, the predictor ***forest cover*** has negative value for accuracy importance (indicating the predictor explains little or nothing), but the proportion of variance explained R² is the highest for that predictor.
Does it make sense to interpret variables with high R² values even if the accuracy of those predictors are zero (or negative)?
I will appreciate any comments.
Paula Eveline Ribeiro D'Anunciação
Mestra em Ecologia e Tecnologia Ambiental (Unifal-MG)
Doutoranda em Ecologia Aplicada (UFLA)
(35) 9114 5125

_______________________________________________
R-sig-ecology mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology