Hi, everybody I am trying to use the regression trees to map some soil attributes of tropical hillslope areas with rpart package. The regression trees shows good results (in r2 evaluation = 0.28 - it is not so bad for soil attributes mapping comparing wiht others models like "lm" or geostats), but I donĀ“t understand the output of "xpred.rpart" command. This is the result: ###
> printcp(silt3.fit) Regression tree: rpart(formula = silt.3 ~ MDE15FILLE + MID.SLOPE_, data = dataset, method = "anova") Variables actually used in tree construction: [1] MDE15FILLE MID.SLOPE_ Root node error: 263226/133 = 1979.1 n= 133 CP nsplit rel error xerror xstd 1 0.095908 0 1.00000 1.02232 0.15190 2 0.040496 1 0.90409 0.99901 0.14677 3 0.028707 2 0.86360 1.06906 0.15292 4 0.028383 4 0.80618 1.03300 0.15267 5 0.023571 5 0.77780 1.02786 0.14994 6 0.018586 6 0.75423 1.03126 0.13944 7 0.012423 7 0.73564 1.05174 0.14205 8 0.010000 8 0.72322 1.09176 0.15294 > xpred.rpart(silt3.fit)[1:10,] 0.54795416 0.06232064 0.03409537 0.02854415 > 0.02586509 0.02093086 0.01519568 0.01114606 1 148.0317 138.8494 138.8494 147.2283 159.0290 159.0290 159.0290 174.1000 2 149.7717 141.1080 141.1080 141.1080 141.1080 151.8475 178.4000 178.4000 14 146.0733 168.6925 152.7095 152.7095 152.7095 152.7095 152.7095 160.8357 15 148.3492 140.4056 140.4056 140.6931 140.6931 140.6931 140.6931 140.6931 23 148.3492 171.1548 153.1118 153.1118 153.1118 153.1118 153.1118 153.1118 39 148.5101 170.6806 149.9600 149.9600 149.9600 149.9600 149.9600 149.9600 52 148.5101 170.6806 149.9600 149.9600 149.9600 149.9600 149.9600 149.9600 53 148.3492 140.4056 140.4056 140.6931 140.6931 140.6931 140.6931 140.6931 56 148.3492 140.4056 140.4056 152.8780 158.6294 158.6294 158.6294 158.6294 57 146.0733 168.6925 152.7095 152.7095 152.7095 152.7095 152.7095 136.4571 #### I would like to know if each column is the cross validation for each variable? And what type of cross validation is use? Leave-one-out? the xval is not set. The CP - complexity parameter, for each column, is different from the "printcp" command, why ? how can I calculate the R2 of the cross validation? or, is the "xstd" column of "printcp" comand teh cross validation? thanks for any help -- ________________________ Waldir de Carvalho Junior Pedologia Pesquisador Embrapa Solos [[alternative HTML version deleted]]
_______________________________________________ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo