Deck 8: Cluster Analysis

Question 1

Which one of the following is NOT an optimization method?

A) determining similarity (or dissimilarity) between units in a cluster relative to units in other clusters
B) examining standardized coefficients
C) maximizing the distance between clusters
D) minimizing within-clusters variation

Answer 1

B
Explanation: Optimization could be based on minimizing the variation within-clusters, maximizing the distance between clusters, or determining similarity (or dissimilarity) between units in a cluster relative to units in other clusters.

Question 2

A researcher concerned with outliers in the clustering model should consider which one of the following cluster algorithms?

A) average linkage
B) complete linkage
C) single linkage
D) none of the above

Answer 2

A
Explanation: As compared to single and complete linkage, outliers are less problematic with average linkage given the use of all units.

Question 3

The starting point identified in nonhierarchical clustering algorithms is which one of the following?

A) Chebychev distance
B) cluster seed
C) group centroid
D) random seed

Answer 3

B
Explanation: Nonhierarchical clustering procedures usually begin by identifying a starting point (i.e., cluster seed).

Question 4

A graphical tool that can be used to assist in determining the number of factors to retain is which one of the following?

A) Akaike's Information
B) dendogram
C) Q-Q plot
D) scree plot

Question 5

Which one of the following is true about the agglomeration schedule?

A) Homogeneity of units is depicted by smaller coefficients
B) The coefficients in the agglomeration schedule represent the number of clusters to retain
C) The number of stages in the agglomeration schedule is one plus the sample size
D) The stopping point in the solution is indicated by small differences in coefficient values

Question 6

Which one of the following is true about sample size determination in cluster analysis?

A) Appropriate sample size is reached when the observed probability is greater than alpha
B) Cluster analysis can be reliably applied to sample sizes as small as 10
C) Power analysis can be used to determine sample size
D) Sufficiency in generating representation of the smallest group is the most important indicator for sample size

Question 7

Which one of the following is a recommended approach for a researcher who is working with variables with disparate measurement scales?

A) Collapse the categories to an ordinal response scale
B) Remove the variables that do not have common response scales
C) Select a different statistical procedure
D) Standardize the variables

Question 8

Which one of the following is the furthest neighbor algorithm selection method?

A) average linkage
B) between-groups linkage
C) complete linkage
D) Ward's method

Question 9

Which one of the following algorithms may produce inaccurate results in cases where clusters have small number of cases?

A) average linkage
B) between-groups linkage
C) complete linkage
D) Ward's method

Question 10

Which one of the following is true about nonhierarchical clustering algorithms?

A) The combined clusters are those that minimize the total sum of squared distance
B) They determine cluster similarity using the maximum or farthest distance between units in a cluster
C) They have small and relatively equal within-cluster variation
D) They minimize the within-cluster variation and maximize the between-cluster distance