Question 1

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

Accepted Answer

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

Accepted Answer

All the terms "Terminal node," "Leaf," and "Bucket" can be used to describe an outcome value in a decision tree, as they all refer to the final nodes that represent classification or decision outcomes.

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.

Accepted Answer

Paths through the decision tree lead to terminal nodes where the outcome variable is predicted based on the values of the classifier variables associated with that node. Therefore, paths through the tree represent constellations of classifier variable values associated with the outcome.

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 ________________.

Accepted Answer

The answer of In a classification tree, there may be...

Question 5

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

Accepted Answer

For example, tree rules may say "If above $1 million, then outcome A. If below $1 million, then if below $250,000 then outcome B, else outcome C." This is true of both classification and regression trees.

Question 6

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

Accepted Answer

The answer of The label for the first or top...

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.

Accepted Answer

Decision trees may be used in the exploratory phase of analysis, prior to using a more traditional confirmatory exploratory procedure.

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.

Accepted Answer

This is one reason why predictions might be better with decision trees.

Question 9

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

Accepted Answer

Predictor variables in decision tree analysis can 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

Accepted Answer

These are other reasons why decision tree predictions may outperform linear regression. Linear regression performs best when its assumptions are met, including linearity, absence of multicollinearity, absence of heteroscedasticity, and having a continuous-normal outcome variable.

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.

Accepted Answer

Categorical variables, either multinomial or ordinal, may be directly entered as predictors. Depending on the specific tree software used, variables may even be of character type.

Question 12

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

Accepted Answer

All analytic procedures work best when model specification includes the true causes and omits spurious causes. Decision trees will use only a subset of predictor variables entered but the variables entered may not be good ones. The "Garbage in, garbage out" rule applies to all analytic procedures, whether decision trees, linear regression, or any other technique.

Question 13

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

Accepted Answer

The answer of Name the most common approach used to...

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.

Accepted Answer

Answer A is not an accepted standard, though some might use this rule of thumb. Answer B is false because, except for conditional inference trees like "ctree", significance testing is not used in decision tree analysis. Since Answer C is true, Answer D is false.

Question 15

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

Accepted Answer

The answer of The process of simplifying a decision tree...

Quiz 4: Classification and Regression Trees in R

What are the two major types of decision tree? _ and ____ trees.

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

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