{"id":2028,"date":"2017-07-15T10:04:32","date_gmt":"2017-07-15T09:04:32","guid":{"rendered":"http:\/\/pcool.dyndns.org:8080\/statsbook\/?page_id=2028"},"modified":"2025-07-01T20:40:34","modified_gmt":"2025-07-01T19:40:34","slug":"decision-tree-plot","status":"publish","type":"page","link":"https:\/\/pcool.dyndns.org\/index.php\/decision-tree-plot\/","title":{"rendered":"Decision Tree Plot"},"content":{"rendered":"\n

A decision tree plot can be useful to visualise decision rules for categorical or continuous outcome variables. A popular package to create decision trees in R is rpart1<\/a><\/sup>. In addition, there is a separate package to produce plots with rpart; rpart.plot2<\/a><\/sup>. Both packages should be installed<\/a>.<\/p>\n\n\n\n

Binary (categorical) outcome<\/strong><\/p>\n\n\n\n

The rpart package\u00a03<\/a><\/sup> contains a data frame with information about patients with kyphosis. To view the first 6 observations:<\/p>\n\n\n\n

library(rpart)\nhead(kyphosis)<\/mark>\n  Kyphosis Age Number Start\n1   absent  71      3     5\n2   absent 158      3    14\n3  present 128      4     5\n4   absent   2      5     1\n5   absent   1      4    15\n6   absent   1      2    16<\/mark><\/em><\/code><\/pre>\n\n\n\n
The data frame contains 4 variables (str stands for structure):<\/p>\n\n\n\n
str(kyphosis)<\/em><\/span>\n'data.frame': 81 obs. of 4 variables:<\/em><\/span>\n $ Kyphosis: Factor w\/ 2 levels \"absent\",\"present\": 1 1 2 1 1 1 1 1 1 2 ...<\/em><\/span>\n $ Age : int 71 158 128 2 1 1 61 37 113 59 ...<\/em><\/span>\n $ Number : int 3 3 4 5 4 2 2 3 2 6 ...<\/em><\/span>\n $ Start : int 5 14 5 1 15 16 17 16 16 12 ...<\/em><\/span><\/code><\/pre>\n\n\n\n
The variable ‘Kyphosis’ is categorical and is either ‘absent’ or ‘present’. The variable ‘Number’ is integer and shows the number of vertebrae involved. The variable ‘Start’ is the highest vertebra.<\/p>\n\n\n\n
To create a decision tree (machine learning) plot is straight forward:<\/p>\n\n\n\n
model <- rpart(Kyphosis ~ Age + Number + Start, data = kyphosis, method = 'class')<\/em><\/span><\/code><\/pre>\n\n\n\n
The variable to the left of the  tilde (~) is the outcome (response) variable and the variables to the right of the tilde are the explanatory variables. To show the model:<\/p>\n\n\n\n
model<\/mark>\nn= 81 \n\nnode), split, n, loss, yval, (yprob)\n      * denotes terminal node\n\n 1) root 81 17 absent (0.79012346 0.20987654)  \n   2) Start>=8.5 62  6 absent (0.90322581 0.09677419)  \n     4) Start>=14.5 29  0 absent (1.00000000 0.00000000) *\n     5) Start< 14.5 33  6 absent (0.81818182 0.18181818)  \n      10) Age< 55 12  0 absent (1.00000000 0.00000000) *\n      11) Age>=55 21  6 absent (0.71428571 0.28571429)  \n        22) Age>=111 14  2 absent (0.85714286 0.14285714) *\n        23) Age< 111 7  3 present (0.42857143 0.57142857) *\n   3) Start< 8.5 19  8 present (0.42105263 0.57894737) *<\/mark><\/em><\/code><\/pre>\n\n\n\n
A plot is more informative:<\/p>\n\n\n\n
plot(model)<\/em><\/span>\ntext(model)<\/em><\/span><\/code><\/pre>\n\n\n\n
\"\"<\/figure>\n\n\n\n
Although it is possible to reduce the text size (with cex = 0.8), the plot doesn’t look very good. However, there is an additional package rpart.plot4<\/a><\/sup> for enhanced graphics.<\/p>\n\n\n\n
library(rpart.plot)<\/span><\/em>\nrpart.plot(model)<\/span><\/em><\/code><\/pre>\n\n\n\n
\"\"<\/figure>\n\n\n\n
Continuous outcome variable (regression)<\/strong><\/p>\n\n\n\n
The rpart package also contains a data frame cu.summary with information about different cars. To show the data frame’s first six observations as well as its structure:<\/p>\n\n\n\n
library(rpart)<\/em><\/span>\nhead(cu.summary)<\/em><\/span>\n                Price Country Reliability Mileage  Type\nAcura Integra 4 11950   Japan Much better      NA Small\nDodge Colt 4     6851   Japan        <NA>      NA Small\nDodge Omni 4     6995     USA  Much worse      NA Small\nEagle Summit 4   8895     USA      better      33 Small\nFord Escort   4  7402     USA       worse      33 Small\nFord Festiva 4   6319   Korea      better      37 Small<\/mark><\/em>\n\nstr(cu.summary)<\/em><\/span>\n'data.frame':\t117 obs. of  5 variables:\n $ Price      : num  11950 6851 6995 8895 7402 ...\n $ Country    : Factor w\/ 10 levels \"Brazil\",\"England\",..: 5 5 10 10 10 7 5 6 6 7 ...\n $ Reliability: Ord.factor w\/ 5 levels \"Much worse\"<\"worse\"<..: 5 NA 1 4 2 4 NA 5 5 2 ...\n $ Mileage    : num  NA NA NA 33 33 37 NA NA 32 NA ...\n $ Type       : Factor w\/ 6 levels \"Compact\",\"Large\",..: 4 4 4 4 4 4 4 4 4 4 ...<\/mark><\/em><\/code><\/pre>\n\n\n\n
To predict the car’s mileage from Price, Country, Reliability and Type (using analysis of variance: anova):<\/p>\n\n\n\n
model2 <- rpart(Mileage ~ Price + Country + Reliability + Type, method=\"anova\", data=cu.summary)<\/span><\/em>\nplot(model2)<\/span><\/em>\ntext(model2)<\/span><\/em><\/code><\/pre>\n\n\n\n
\"\"<\/figure>\n\n\n\n
Or for a nicer plot:<\/p>\n\n\n\n
library(rpart.plot)<\/em><\/span>\nrpart.plot(model2)<\/em><\/span><\/code><\/pre>\n\n\n\n
\"\"<\/figure>\n","protected":false},"excerpt":{"rendered":"
A decision tree plot can be useful to visualise decision rules for categorical or continuous outcome variables. A popular package to create decision trees in R is rpart. In addition, there is a separate package to produce plots with rpart; rpart.plot. Both packages should be installed. Binary (categorical) outcome The rpart package\u00a0 contains a data […]<\/p>\n","protected":false},"author":2,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"inline_featured_image":false,"footnotes":""},"class_list":["post-2028","page","type-page","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/pcool.dyndns.org\/index.php\/wp-json\/wp\/v2\/pages\/2028","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/pcool.dyndns.org\/index.php\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/pcool.dyndns.org\/index.php\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/pcool.dyndns.org\/index.php\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/pcool.dyndns.org\/index.php\/wp-json\/wp\/v2\/comments?post=2028"}],"version-history":[{"count":1,"href":"https:\/\/pcool.dyndns.org\/index.php\/wp-json\/wp\/v2\/pages\/2028\/revisions"}],"predecessor-version":[{"id":4736,"href":"https:\/\/pcool.dyndns.org\/index.php\/wp-json\/wp\/v2\/pages\/2028\/revisions\/4736"}],"wp:attachment":[{"href":"https:\/\/pcool.dyndns.org\/index.php\/wp-json\/wp\/v2\/media?parent=2028"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}