Regression tree in R

Question

I am having trouble making a regression tree in R. I have a data frame with 17 attributes

library(rpart)
rt.model <- rpart(razlika ~ ., learn)

I get an error:

Error in `[.data.frame`(frame, predictors) : undefined columns selected

Seems weird because I did something like that with a very simillar database. You can dowload the dataframe on http://uploading.com/files/de8a966d/exa.Rda/ - then load with

load("exa.Rda")

Answer 1

The problem is not, I believe, that you have a matrix rather than a data frame. When I download and then load you data set, I get a data frame, not a matrix.

The problem is that you have bad characters in the column names. Use gsub to remove the characters "-", " ", "(" and ")" from the column names. Or you can simply redefine the column names yourself entirely using colnames.

Or do as ulvund does and simply call data.frame, which forces R to do the column name cleaning for you, by default.

When I do this, rpart runs just fine.

Answer 2

Turn your learn matrix into a data frame.

Example:

load("exa.Rda")
library(rpart)
learn <- data.frame(learn)
rt.model <- rpart(razlika ~ ., learn)
rt.model

yields:

n= 226 

node), split, n, deviance, yval
      * denotes terminal node

  1) root 226 31417.5100   3.3849560  
    2) B.reb>=40.80799 117 12661.2300   0.4871795  
      4) B.ft>=0.7666193 31  2685.4190  -5.7741940  
        8) A.fg2< 0.4645683 22  1846.7730  -8.3181820  
         16) A.ft< 0.7464692 7   365.4286 -14.2857100 *
         17) A.ft>=0.7464692 15  1115.7330  -5.5333330 *
        9) A.fg2>=0.4645683 9   348.2222   0.4444444 *
      5) B.ft< 0.7666193 86  8322.3720   2.7441860  
       10) B.avg.conceded.< 98.19592 76  7255.6320   1.7105260  
         20) A.reb< 39.29941 19  1520.6320  -3.5789470 *
         21) A.reb>=39.29941 57  5026.2110   3.4736840  
           42) A.3pt< 0.3945418 35  2500.1710   0.7714286  
             84) A.ft< 0.7460665 17  1270.2350  -2.4705880 *
             85) A.ft>=0.7460665 18   882.5000   3.8333330 *
           43) A.3pt>=0.3945418 22  1863.8640   7.7727270  
             86) B.ft>=0.7214165 13   718.9231   4.0769230 *
             87) B.ft< 0.7214165 9   710.8889  13.1111100 *
       11) B.avg.conceded.>=98.19592 10   368.4000  10.6000000 *
    3) B.reb< 40.80799 109 16719.2500   6.4954130  
      6) A.fouls>=24.51786 23  2349.9130  -2.2173910  
       12) A.fg2< 0.4551468 16  1266.0000  -5.5000000 *
       13) A.fg2>=0.4551468 7   517.4286   5.2857140 *
      7) A.fouls< 24.51786 86 12156.3800   8.8255810  
       14) B.fouls< 22.80863 24  3271.9580   2.5416670  
         28) A.3pt< 0.3738479 9   626.0000  -6.0000000 *
         29) A.3pt>=0.3738479 15  1595.3330   7.6666670 *
       15) B.fouls>=22.80863 62  7569.8710  11.2580600  
         30) A.fouls< 22.32999 18  1650.5000   5.5000000 *
         31) A.fouls>=22.32999 44  5078.4320  13.6136400  
           62) A.ft.drawn>=29.18849 7   208.8571   3.8571430 *
           63) A.ft.drawn< 29.18849 37  4077.1890  15.4594600  
            126) A.fg2< 0.4588535 18  1696.5000  11.5000000 *
            127) A.fg2>=0.4588535 19  1831.1580  19.2105300 *

Answer 3

This can also happen if column names are integers (1:N), even though they are stored as characters.