Lecture 15 - Non-Parametric Tests Part 2 Flashcards
In what situation is a Kruskal-Wallis H test suitable?
When the data is non-parametric, the design is between-groups, there is one IV with multiple levels and the DV is ordinal.
What is the rationale behind a Kruskal-Wallis H test?
Distribution of ranks across groups is either random or systematic depending on whether H1 is true.
How is a Kruskal-Wallis H test carried out by hand?
All scores are ranked ignoring the IV.
H is calculated using N, the sum of ranks for each level of the variable (Ri) and the number of scores in each level of the variable (ni).
How is a Kruskal-Wallis H test carried out in SPSS?
Analyse/Non-parametric tests/k independent samples.
Test variable= DV, grouping variable=IV. Define range and check Kruskal-Wallis box.
In SPSS, what output is there for a Kruskal-Wallis H test and what is the useful information?
Ranks and Test Statistics.
SPSS converts H into X2, and the df and sig. are also useful.
What is a problem with Kruskal-Wallis H tests and how can this be corrected for?
They tell us that the conditions differ, but not which ones, so to find this out multiple Mann-Whitney U tests with Bonferroni corrections must be carried out.
In what situation should a Friedman test be used?
When the data are non-parametric, the samples are paired (within-subject and matched-subject designs), there is one IV with >2 levels (3+) and the DV is an ordinal scale.
What is the rationale behind a Friedman test?
Each individual subject’s scores are ranked (i.e. for three IV levels the ranks are 1-3) and the rank totals for each IV level calculated. If H0=true, expect random rank distribution and roughly equal totals.
How is a Friedman test carried out?
Once the IV level subject ranks have been totalled (R), the test statistic, X2R, is calculated using N, R and k (no. of IV levels).
How do you conduct a Friedman test on SPSS?
Analyse/Nonparametric Tests/k Related Samples
Test variables = IV levels
Check Friedman box.
What does the output from a Friedman test in SPSS contain and what information is useful?
Ranks and Test Statistics.
X2, df and sig. values are reported.
What is a problem with Friedman tests and how can it be solved?
As with Kruskal-Wallis, it doesn’t say which conditions differ. Multiple Bonferroni-corrected Wilcoxon tests can be used to determine this.