Deck 4: Classification and Regression Trees in R

Question 1

What are the two major types of decision tree? _______________________ and __________________________ trees.

Answer 1

Classification and regression trees

Question 2

In a decision tree, what is a label for one of the outcome values?

A) Terminal node
B) Leaf
C) Bucket
D) All of the above

Answer 2

D

Question 3

In a decision tree, paths through the tree to an outcome value represent:

A) Causal sequences
B) Constellations of classifier variable values associated with the terminal node outcome
C) Independent variables in descending order of correlation with the outcome variable.
D) Independent variables in ascending order of correlation with the outcome variable.

Answer 3

B

Question 4

In a classification tree, there may be more than one path to the same terminal value of the outcome variable. This attribute of classification trees is labeled ________________.

Question 5

The same predictor variable may appear in different paths in a decision tree and even more than once in the same path.

Question 6

The label for the first or top node in a decision tree is the ______________ node.

Question 7

What is a way decision tree analysis might complement OLS regression analysis?

A) Model specification, to select predictor variables
B) Reveal unobserved heterogeneity (subpopulations requiring different predictors)
C) Identify optimal cutting points when binning continuous variables.
D) All of the above.

Question 8

Interaction effects are handled automatically by decision trees, so there is no need to add explicit interaction terms as one would in OLS regression.

Question 9

In decision tree analysis, predictor variables may be nominal, ordinal, or continuous in level.

Question 10

Decision trees are

A) Nonparametric
B) Nonlinear
C) Robust against multicollinearity
D) All of the above

Question 11

In decision tree analysis, it is not necessary to create dummy variables for categorical variables or to interpret results in terms of reference categories.

Question 12

Selecting salient variables prior to running the analysis is more important in decision trees than in OLS regression.

Question 13

Name the most common approach used to address the problem of overfitting and failure to generalize in decision tree analysis? ____________________________________

Question 14

What is true of sample size in relation to decision tree analysis?

A) Ten times the number of predictor variables is the minimum number of observations required for decision tree solutions.
B) Significance testing of decision tree solutions is unreliable with small samples.
C) In larger trees, lower nodes may be based on too few observations, meaning the solution will lack statistical power.
D) Sample size does not matter in decision tree analysis.

Question 15

The process of simplifying a decision tree solution by removing nodes is called __________________.