Ch.20 Witte Tests For Ranked Data Verified Test Bank

MULTIPLE‑CHOICE TEST ITEMS

CHAPTER 20

TESTS FOR RANKED DATA

20.1 Replace the t test for two independent samples with the Mann‑Whitney U test whenever the assumptions of the t test appear to be violated and

a) data are qualitative.

b) a type II error is particularly serious.

c) measurement is inaccurate.

d) sample sizes are small.

20.2 As is true of all tests for ranked data, the U test is immune to violations of assumptions about

a) normality and equal variances.

b) the sampling distribution.

c) random samples.

d) the type of data.

20.3 Strictly speaking, the null hypothesis for the U test equates two population

a) means.

b) medians.

c) shapes.

d) distributions.

20.4 When converting a list of original scores into ranks, list all observations

a) without regard to order for each separate group.

b) without regard to order for the combined groups.

c) from smallest to largest for each separate group.

d) from smallest to largest for the combined groups.

20.5 When converting original scores to ranks, if three original scores have identical values, they should be

a) discarded.

b) arbitrarily assigned the three ranks that would have been used had the scores been slightly different.

c) assigned the median of the ranks that would have been used had the scores been slightly different.

d) assigned the single rank that would have been used had there only been one original score.

20.6 The more one group tends to outrank the other group, the

a) more suspect the null hypothesis.

b) less suspect the null hypothesis.

c) more suspect the original assumptions.

d) less suspect the original assumptions.

20.7 An unusual feature of hypothesis tests involving U is that

a) there is no single sampling distribution for U.

b) only nondirectional tests are possible.

c) the null hypothesis is rejected if the observed U is less than or equal to the critical U.

d) the decision to retain the null hypothesis is stronger than the decision to reject the null hypothesis.

20.8 Given an exclusive concern about a population difference in a particular direction, a directional U test is

a) never appropriate.

b) appropriate.

c) appropriate, assuming that the population distributions are known.

d) appropriate, assuming that the population distributions have roughly similar variabilities and shapes.

20.9 The Wilcoxon T test is designed to replace the t test for two related samples whenever sample sizes are

a) small.

b) small and the normality assumption is suspect.

c) unequal.

d) unequal and the normality assumption is suspect.

20.10 When ordering difference scores from smallest to largest (prior to calculating Wilcoxon's T), ignore all

a) difference scores of zero.

b) negative difference scores.

c) tied difference scores.

d) extreme difference scores.

20.11 If the value of Wilcoxon's T equals zero,

a) a computational mistake has been made.

b) all of the original difference scores have the same sign.

c) the original difference scores are equally split in terms of positive and negative signs.

d) the null hypothesis should be retained.

20.12 The null hypothesis for the Kruskal‑Wallis H test specifies the equality of the various population

a) means.

b) medians.

c) distributions.

d) shapes.

20.13 The Kruskal‑Wallis H test is designed to replace the

a) one-factor F test.

b) two-factor F test.

c) two-variable chi-square test.

d) any t test.

20.14 When sample sizes are not extremely small, critical values for the H test can be obtained from the

a) normal tables.

b) t tables.

c) F tables.

d) chi‑square tables.

20.15 Compared to both the U test and the T test, a distinguishing characteristic of the H test is its

a) more conventional decision rule.

b) vague null hypothesis.

c) use of approximate critical values.

d) treatment of tied scores.

20.16 Refer to more advanced books for a possible adjustment if there are more than a few ties and the observed value of U, T, or H is

a) very small.

b) very large.

c) in the vicinity of its critical value but not statistically significant.

d) just slightly beyond its critical value.

20.17 Strictly speaking, all tests in this chapter assume

a) normal populations.

b) no ties in ranks.

c) equal population variances.

d) all of the above.

20.18 Designating the U, T and H tests as nonparametric tests refers to the notion that these tests are not based on

a) well‑defined sampling distributions.

b) conventional hypothesis testing procedures.

c) null hypotheses for specific population characteristics.

d) null hypotheses for general population characteristics.

20.19 Insofar as the U, T, and H tests can be conducted in the absence of assumptions about the population distributions, these tests can be referred to as

a) nonparametric.

b) parametric.

c) distribution‑free.

d) distribution‑bound.

20.20 Insofar as population distributions are assumed to have roughly similar variabilities and shapes, the U, T, and H tests qualify as

a) neither nonparametric nor distribution-free.

b) nonparametric only.

c) distribution-free only.

d) both nonparametric and distribution-free.

20.21 When data are quantitative and populations appear to be normally distributed with equal variances, use the

a) U, T, and H tests.

b) t and F tests.

c) z test.

d) chi‑square test.

20.22 When used inappropriately as replacements for the t and F tests, the U, T, and H tests are more likely to cause the

a) retention of a true null hypothesis.

b) retention of a false null hypothesis.

c) rejection of a true null hypothesis.

d) rejection of a false null hypothesis.

20.23 Beware of non-normality when sample sizes are

a) large.

b) moderate.

c) small (less than about 20).

d) small (less than about 10).

20.24 Beware of unequal variances when sample sizes are

a) small and equal.

b) small and unequal.

c) large and equal.

d) large and unequal.

20.25 When data are quantitative and all assumptions appear to be satisfied, the t and F tests are preferred to the U, T, and H tests because the former tests

a) minimize the probability of a type I error.

b) minimize the probability of a type II error.

c) are more readily understood.

d) are more readily calculated.