Kruskal-Wallis H Test for Independent Samples & Friedman Test for Repeated Measures

Definition (#f7aeae)

Important (#edcae9)

Extra (#fffe9d)

Kruskal-Wallis H Test for Independent Samples:

Characteristics:

  • DV can be ordinal, interval, or ratio.

  • IV is categorical/nominal with more 2 levels/groups.

  • Test differences between conditions/groups.

Concept:

  • Similar to Mann-Whitney test, Kruskal-Wallis test analyses data using ranking.

  • Involves arranging data in ascending order, and totaling the rank for each group.

  • If scores of groups are clustered at either end of sequence, reflects a significant group difference.

  • If scores of groups are randomly distributed throughout, reflects no significant group difference.

  • Ex:

    • Amount of hours spent watching favourite genre of tv shows per week.

Analysis:

  1. Analyze → Nonparametric Tests→ Legacy Dialogues → k Independent Samples

  2. Put IV variable in “Grouping Variable” and DV in “Test Variable List”.

    Step 2

  3. Tick “Kruskal-Wallis H” → Define Range for IV. Click “Continue” → Click “OK”.

Step 3Output

Further Analysis:

  • Significant Kruskal-Wallis test shows group difference, but not which group have difference.

  • If significant, Mann-Whitney tests must be done to do pairwise comparison between each group.

  • Due to possible family-wise error, Bonferroni correction must be done.

  • New alpha value = 0.05 / number of Mann-Whitney test.

Template:

  • What test you ran & what variables were plugged in.

  • Is there a significant difference between ranking.

  • Report the “equation” with the values.

  • Interpretation:

    • If significant: 1 group has higher score than the other.

    • If not significant: Similar scores across groups.

Friedman Test for Repeated Measures:

Characteristics:

  • Within-subject test.

  • IV has more than two levels.

  • Variables must be ordinal, scale, or ratio.

Concept:

  • Rank each participant’s score.

    • Within participants across the levels.

  • Test the difference between the ranking of the levels.

    • Whether there is a systematic distribution and random distribution of ranking.

  • Ex:

    • Total amount of hours spent specific genre of tv shows per week.

Analysis:

  1. Analyze → Nonparametric Tests→ Legacy Dialogues → k Related Samples.

  2. Put variables in “Test Variables” → Tick “Friedman” → Click “OK”.

    Step 2
Output

Further analysis:

  • Significant Friedman test for repeated measures shows level difference, but not which level have difference.

  • If significant, Wilcoxon tests must be done to do pairwise comparison between each level.

  • Due to possible family-wise error, Bonferroni correction must be done.

  • New alpha value = 0.05 / number of Wilcoxon test

Template:

  • What test you ran & what variables were plugged in.

  • Is there a significant difference between ranking.

  • Report the “equation” with the values.

  • Interpretation:

    • If significant: 1 level has higher score than the others.

    • If not significant: Similar scores across levels.