Session Ready
Exercise

Evaluating a recursive partitioning model

Consider this model formula about runners' net times: net ~ age + sex. The graphic shows the recursive partitioning for this formula. At the very bottom of the tree, in circles, are values for the response variable. At the very top is the root. (Modeling convention is to draw such trees upside down compared to the familiar botantical form, where the roots are at the bottom.)

Training an rpart() model amounts to finding a set of divisions of the cases. Starting with all the cases at the root, the model divides them up into two groups: males on the left and females on the right. For males, a further split is made based on age: those younger than 50 and those 50 and over. Similarly, females are also split on age, with a cut-point of 46 years. So, for a 40 year-old female, the model output is 93 (with the same units as the response variable: minutes).

The rpart() function uses a sensible default for when to stop dividing subgroups. You can exercise some control over this with the cp argument.

Using the console, train a model with the same formula as used in the graphic, but with a value of cp = 0.001. Display the model as a tree. Your commands will look like this:

model_2 <- rpart(___, data = Runners, cp = 0.001)
prp(model_2, type = 3)

The prp() function plots the model as a tree. type = 3 is one of several available formats. In this model (with cp = 0.001), what is the model output for a 58 year-old female?

Instructions
50 XP
Possible Answers