Rpart Cross Validation

["Annette Molinaro" <molinaro@cc.ucsf.edu>]

Greetings.

 

I am running Rpart in R 1.6.1  - when looking at a datset with n=1000, minsplit=20=mins,method="anova", and using weights (ranging from 0 to 5) I get the following:

    Regression tree:
    rpart(formula = c50.time ~ a1$x1 + a1$x2 + a1$x3 + a1$x4 + a1$x5,
        weights = c50.pi.wt, method = "anova", xval = 5, minsplit = mins)

 

    Variables actually used in tree construction:
    [1] a1$x1 a1$x2

 

    Root node error: 136.97/1000 = 0.13697

 

    n= 1000 

 

            CP nsplit rel error xerror     xstd
    1 0.068713      0   1.00000 1.0042 0.044093
    2 0.039362      1   0.93129 1.0042 0.044093
    3 0.010000      2   0.89192 1.0042 0.044093 

 

Question : How is it possible to get the exact same cross validation error and standard deviation for all three trees?

 

Thanks,

Annette