Deck 16: Non-Parametrics Revisited: the Ordinal Case

Question 1

The ordinal test for assessing the difference among three or more correlated samples, is the

A) Wilcoxon T
B) Mann-Whitney U
C) Kruskal-Wallis H
D) Friedman ANOVA by ranks

Answer 1

Friedman ANOVA by ranks

Question 2

In order to calculate an ordinal test of significance, the following must be true

A) the scores must be normally distributed in the population
B) the variability within each sample must be homogeneous
C) the population means must be assumed to be equal
D) none of these

Answer 2

D

Question 3

The Pearson r is to the Spearman rs , as the

A) F is to t
B) chi square is to V
C) H is to U
D) independent F is to H

Answer 3

independent F is to H

Question 4

In terms of the amount of information being contained in the measures, the following measurement scales rank

A) first nominal, then ordinal, and then interval
B) first ordinal, then nominal, and then interval
C) first interval, then nominal, and then ordinal
D) first interval, then ordinal, and then nominal

Question 5

The Kruskal-Wallis H is the ordinal equivalent of which interval test?

A) independent t
B) paired t
C) independent F
D) within-subjects ANOVA

Question 6

The ordinal test for assessing the difference between two independent samples, is the

A) Wilcoxon T
B) Mann-Whitney U
C) Kruskal-Wallis H
D) Spearman r

Question 7

The ordinal test for assessing the difference between three or more independent samples, is the

A) Wilcoxon T
B) Mann-Whitney U
C) Kruskal-Wallis H
D) Spearman r

Question 8

The Mann-Whitney U is the ordinal equivalent of which interval test?

A) independent t
B) paired t
C) independent F
D) within-subjects ANOVA

Question 9

The most appropriate measure of central tendency on ordinal data is the

A) mean
B) median
C) mode
D) interquartile range

Question 10

The Wilcoxon T test is the ordinal equivalent of which interval test?

A) independent t
B) paired t
C) independent F
D) within-subjects ANOVA

Question 11

When the conversion is appropriate, it is always possible to convert

A) nominal into interval data
B) ordinal into interval data
C) nominal into ordinal data
D) interval into ordinal data

Question 12

The U test was created by

A) Wilcoxon
B) Mann and Whitney
C) Kruskal and Wallis
D) Friedman

Question 13

Of the following correlation coefficients, which requires ordinal data?

A) the Pearson r
B) the coefficient of contingency
C) the Spearman rs
D) all of these

Question 14

With the ordinal data, the important information being provided concerns

A) equality versus non-equality
B) greater-than, or less-than
C) the distance between the successive scale points
D) an absolute zero

Question 15

The T for correlated samples was created by

A) Wilcoxon
B) Mann and Whitney
C) Kruskal and Wallis
D) Friedman

Question 16

The Friedman ANOVA by ranks is the ordinal equivalent of which interval test?

A) independent t
B) paired t
C) independent F
D) within-subjects ANOVA

Question 17

Ordinal tests of significance are all

A) parametrics
B) non-parametrics
C) distribution free
D) both b and c, but not a

Question 18

The H test was created by

A) Wilcoxon
B) Mann and Whitney
C) Kruskal and Wallis
D) Friedman

Question 19

The ANOVA by ranks was created by

A) Wilcoxon
B) Mann and Whitney
C) Kruskal and Wallis
D) Friedman

Question 20

The ordinal test for assessing the difference between two correlated samples is the

A) Wilcoxon T
B) Mann-Whitney U
C) Kruskal-Wallis H
D) Spearman r

Question 21

Establishing whether the difference between two independently selected sample means is significant, requires the

A) U test
B) H test
C) T test
D) none of these

Question 22

For the ordinal T on small samples, with an obtained value of 17, and a critical, tabled value of 6.00, then

A) Ho is rejected
B) There is a failure to reject Ho
C) Ho cannot be tested
D) the sample mean equals the parameter mean

Question 23

If ordinal data were obtained from a three-group, matched-subjects design, then the appropriate test would be the

A) T
B) H
C) U
D) Friedman ANOVA by ranks

Question 24

The ordinal test, cited in the text, for testing the hypothesis of association is the

A) Pearson r
B) H
C) T
D) Spearman r

Question 25

The ordinal T test shown in the text requires at least

A) 12 pairs of scores
B) 16 independent scores
C) 6 pairs of scores
D) 25 independent scores

Question 26

The Spearman r should only be used to test the hypothesis of

A) difference
B) association
C) equivalence
D) none of these

Question 27

If ordinal data were obtained from a one-group, repeated-measures design, then the appropriate test would be the

A) T
B) H
C) U
D) chi square

Question 28

For the ordinal T on small samples, with a calculated value of 4.13 and a critical, tabled value of 6.00, then

A) Ho is rejected
B) There is a failure to reject Ho
C) Ho cannot be tested
D) the sample mean equals parameter mean

Question 29

Homogeneous variability within the various sample groups is required in order to calculate the

A) U test
B) H test
C) Spearman r
D) none of these

Question 30

If ordinal data were obtained from a four-group, between-subjects design, then the appropriate test would be the

A) T
B) H
C) U
D) Friedman ANOVA by ranks

Question 31

Ordinal tests may be used to test data from the

A) experimental method
B) post-facto method
C) both of these
D) neither of these

Question 32

When all the assumptions of the interval tests are not met, the researcher should use

A) parametric test
B) non-parametric tests
C) tests of association
D) none of these

Question 33

Homoscedasticity is required in order to calculate the

A) U test
B) H test
C) T test
D) none of these

Question 34

Compared to the parametric tests, the non-parametrics have

A) more assumptions
B) fewer assumptions
C) the same number of assumptions
D) all of these, depending on the scale of measurement

Question 35

All non-parametric tests demand

A) nominal data
B) ordinal data
C) interval data
D) none of these

Question 36

Compared to the Pearson r, the Spearman r is a

A) more powerful test
B) less powerful test
C) test of equal power
D) none of these, since power cannot be established for the r

Question 37

If ordinal data are obtained from a two-group, between-subjects design (with random assignment of all subjects), the appropriate test would be

A) U
B) T
C) Spearman r
D) Friedman ANOVA by ranks

Question 38

Of the ordinal tests presented in the text, the only one which demands a reject of Ho when the obtained value is smaller than the critical, tabled value, is the

A) T
B) Spearman r
C) U
D) none of these

Question 39

If the calculated value of zu were 3.15 and the critical, tabled value were 1.96, then Ho would be

A) rejected
B) a failure to reject Ho
C) untestable
D) none of these

Question 40

When doing the ordinal T test, it is important to establish

A) the signed ranks
B) frequency of occurrence
C) the difference between the means
D) all of these

Question 41

Of the following, which does not belong with the others?

A) U test
B) H test
C) Pearson r
D) T test

Question 42

The more powerful a given statistical test is, the more likely it is to

A) increase the Type 2 error
B) detect significance
C) produce non-chance results
D) both b and c, but not a

Question 43

Whenever non-parametrics are used instead of parametrics, and the parametric assumptions are met, then

A) power is increased
B) power is reduced
C) power is eliminated
D) power becomes equal to .95

Question 44

Whenever there is an increased probability of accepting null when null should have been rejected, then

A) the Type 2 error is increased
B) the power is decreased
C) there is a decreased likelihood of obtaining significance
D) all of these

Question 45

If a Spearman r were used on data requiring a Pearson r, then in order to obtain significance

A) a larger sample would be required
B) a smaller sample would suffice
C) the means of the two samples would have to deviate significantly from each other
D) the population mean must be a known value

Question 46

In order to test experimental data via the ordinal T test, the design must be

A) between-subjects with independent assignment of subjects
B) repeated-measures
C) matched-subjects
D) both b and c, but not a

Question 47

Using a non-parametric test, when a parametric test should have been used, tends to

A) reduce the likelihood of rejecting Ho
B) increase the likelihood of rejecting Ho
C) over-estimate the difference between the sample mean and the parameter mean
D) none of these

Question 48

In order to test experimental data via the Friedman (ANOVA by ranks), the design must be

A) between-subjects with independent assignment of subjects
B) repeated-measures
C) matched-subjects
D) both b and c, but not a

Question 49

The assumption of a normal distribution within the underlying population of scores, is required with the

A) Spearman r
B) U test
C) chi square
D) none of these

Question 50

In order to test experimental data via the U test, the design must be

A) between-subjects
B) repeated-measures
C) matched-subjects
D) none of these, since U may not test experimental data

Question 51

Conditions which tend to increase beta, also increase power.

Question 52

Of the following, which may be used for the analysis of post-facto data?

A) Spearman r
B) H
C) U
D) all of these

Question 53

All parametric tests are less powerful than the non-parametric.

Question 54

Regardless of the scale of measurement, the researcher analyzing data from a three-group, between-subjects design, must use the

A) T
B) U
C) H
D) none of these

Question 55

The hypothesis of difference may be tested by using

A) the U test
B) the H test
C) Friedman ANOVA by ranks
D) all of these

Question 56

Of the following, which does not belong with the others?

A) Pearson r
B) Spearman r
C) Coefficient of contingency
D) U test

Question 57

Under which of the following conditions does an increase in beta lead to an increase in power?

A) between-subjects design with ordinal data
B) repeated-measures design with interval data
C) matched-subjects design with nominal data
D) none of these, since the premise is false

Question 58

In order to test experimental data via the H test, the design must be

A) between-subjects
B) repeated-measures
C) matched-subjects
D) both b and c, but not a

Question 59

Since it has so few assumptions, chi square may always be used on ordinal data.

Question 60

For a given sample size, the more powerful a test is, the more it is likely to produce

A) significance
B) a lack of significance
C) a larger beta
D) none of these

Question 61

16-70.For a given sample size, the Spearman r is more powerful than the Pearson r

Question 62

Interval data may always be converted into ordinal data.

Question 63

If a distribution of interval scores is badly skewed, the scores should be ranked and ordinal tests performed.

Question 64

The more power a test has, the more likely it is to produce significant results.

Question 65

Ordinal data may always be converted into interval data.

Question 66

All non-parametric tests are also distribution-free.

Question 67

The Friedman ANOVA by ranks does for ordinal data what the U test does for interval data

Question 68

One should not use the Spearman r on skewed interval data, unless the data could be converted into ordinal form.

Question 69

For the questions below, use the following: A researcher used a one-group, repeated-measure design, and established the following data: Scores Ranked by Rows

-Which statistical test should be used?

Question 70

The Spearman r does for ordinal data what the Pearson r does for interval data.

Question 71

The U test makes no assumptions regarding the population parameters.

Question 72

Ordinal data may never be skewed.

Question 73

Non-parametric tests may never be used on data from post-facto research.

Question 74

Ordinal test require more assumptions than nominal tests.

Question 75

The Type 2 error may only occur when the null hypothesis is accepted.

Question 76

The concept of "greater than" or "less than" is at the heart of the ordinal case.

Question 77

The ordinal T for small samples only rejects null when the obtained value is equal to or less then the tabled value.

Question 78

The Type 2 error is the probability of being wrong whenever the null hypothesis is rejected.

Question 79

For the following distribution of interval scores, indicate, in order, the correct rankings. 799, 522, 515, 515, 515, 515, 515, 500.

Question 80

For the following distribution of interval scores, indicate, in order, the correct rankings. 110, 55, 55, 50, 42, 42, 42, 30, 27