6. Friedman and Kruskal-Wallis

Question 1

What is the Friedman test?

Answer 1

A non-parametric version of the one-way repeated-measures ANOVA.

Question 2

What is the Kruskal-Wallis test?

Answer 2

A non-parametric version of the one-way between-subjects ANOVA.

Question 3

When should you use the Friedman test?

Answer 3

When comparing 3+ related conditions with ordinal or skewed data.

Question 4

When should you use the Kruskal-Wallis test?

Answer 4

When comparing 3+ unrelated groups with ordinal or skewed data.

Question 5

What type of design does the Friedman test use?

Answer 5

Within-subjects design.

Question 6

What type of design does the Kruskal-Wallis test use?

Answer 6

Between-subjects design.

Question 7

What does the Friedman test compare?

Answer 7

The mean ranks across conditions.

Question 8

What does the Kruskal-Wallis test compare?

Answer 8

The rank sums across groups.

Question 9

What is an omnibus test?

Answer 9

A test that identifies overall differences but not where they are.

Question 10

What is the null hypothesis for the Friedman test?

Answer 10

No difference between the related conditions.

Question 11

What is the null hypothesis for the Kruskal-Wallis test?

Answer 11

No difference between the unrelated groups.

Question 12

What is the alternative hypothesis for these tests?

Answer 12

There is a difference between the groups/conditions.

Question 13

What is a post-hoc test?

Answer 13

A follow-up test conducted after a significant omnibus test.

Question 14

What post-hoc test is used with Friedman?

Answer 14

Wilcoxon signed-rank or Conover test.

Question 15

What post-hoc test is used with Kruskal-Wallis?

Answer 15

Wilcoxon rank-sum or Conover test.

Question 16

What is the familywise error rate (FWE)?

Answer 16

The increased risk of a Type I error due to multiple comparisons.

Question 17

What is the Bonferroni correction?

Answer 17

A method that adjusts the alpha level to reduce FWE.

Question 18

How do you apply the Bonferroni correction?

Answer 18

Divide the alpha level by the number of comparisons.

Question 19

What is Type I error?

Answer 19

A false positive, rejecting a true null hypothesis.

Question 20

What is Type II error?

Answer 20

A false negative, failing to reject a false null hypothesis.

Question 21

What is Kendall’s W?

Answer 21

An effect size measure for the Friedman test.

Question 22

What are the benchmarks for Kendall’s W?

Answer 22

0.1 = small, 0.5 = medium, 0.8 = large effect.

Question 23

What is eta squared (η²)?

Answer 23

An effect size measure for the Kruskal-Wallis test.

Question 24

What are the benchmarks for eta squared?

Answer 24

0.01 = small, 0.06 = medium, 0.14 = large effect.

Question 25

What is the benefit of omnibus tests?

Answer 25

They reduce the familywise error rate.

Question 26

What is the downside of omnibus tests?

Answer 26

They don't indicate where the difference lies.

Question 27

When should you run post-hoc tests?

Answer 27

Only after a significant omnibus test.

Question 28

What data types can Friedman and Kruskal-Wallis use?

Answer 28

Ordinal, interval, or ratio data.

Question 29

What does a significant Friedman test indicate?

Answer 29

At least one condition is different from the others.

Question 30

What does a significant Kruskal-Wallis test indicate?

Answer 30

At least one group is different from the others.

Question 31

What does a non-significant Friedman test mean?

Answer 31

No significant difference between conditions.

Question 32

What does a non-significant Kruskal-Wallis test mean?

Answer 32

No significant difference between groups.

Question 33

What is the primary output of the Friedman test?

Answer 33

Chi-square (χ²) statistic.

Question 34

What is the primary output of the Kruskal-Wallis test?

Answer 34

Chi-square (χ²) statistic.

Question 35

What is counterbalancing in within-subjects designs?

Answer 35

Rotating the order of conditions to prevent order effects.

Question 36

What is the effect of multiple tests without correction?

Answer 36

Increased Type I error rate.

Question 37

What is the difference between planned and post-hoc comparisons?

Answer 37

Planned comparisons are decided before data collection; post-hoc are done after.

Question 38

Why are effect sizes important in these tests?

Answer 38

They indicate the magnitude of the difference.