June 2017
Beginner to intermediate
576 pages
15h 22m
English
The ctree function gives some advantages over rpart in that the results produced can be a bit more intuitive. Rather than optimizing a result node based purely upon the purity of the resultant node, ctree uses statistical hypothesis testing to determine if the results of the split are statistically significant. It uses chi-square test statistics to test the association, only keeping the associations that are significant, and thus removing bias due to a large number of categories. So, while accuracy may suffer in some cases, benefit is gained by having the results be more explanatory:
install.packages("partykit")
library(partykit)
y2 <- ctree(Life.Exp ~ .,data=x)
y2
plot(y2)
As you can see from the following plot, ctree ...