Non-parametric tests

Question 1

Parametric test

Answer 1

-Based on parameters (e.g. standard deviation)
-Assumes normal distribution of data
-Most common stats tests covered so far are parametric

Parametric tests rely on assumptions about the population parameters and distribution of the data.

Question 2

When not to use parametric test

Answer 2

-When data is not normally distributed
-When scales are ordinal rather than interval or ratio
-When data is skewed or step sizes between values are unequal

These conditions can lead to inaccurate results if parametric tests are applied.

Question 3

Non-parametric test options

Answer 3

-These tests use ranking instead of raw values
-Are robust against non-normal data
-Mann-Whitney U
-Wilcoxon signed-rank
-Kruskal-Wallis H
-Friedman

Non-parametric tests are useful when data does not meet the assumptions required for parametric testing.

Question 4

Mann-Whitney U

Answer 4

-Equivalent to independent t-test
-Use for 2 unpaired samples (between-subjects)
-Null hypothesis- group ranks should be randomly distributed

Use for 2 unpaired samples (between-subject designs).

Question 5

Wilcoxon signed-rank test

Answer 5

-Equivalent to paired samples (within-subject designs)
-Calculates ranked differences between paired observations
-Tests whether positive and negative ranks balance

This test calculates ranked differences between paired observations.

Question 6

Kruskal-Wallis H test

Answer 6

-Equivalent to one-way ANOVA (between-subjects)
-Use for 3+ independent groups
-Null hypothesis- group ranks are equal
-SPSS reports as χ²

Use for 3+ independent groups.

Question 7

Friedman Test

Answer 7

-Equivalent test repeated measures ANOVA
-Use for 3+ related samples (within-subjects)
-Ranks data within each subject

Use for 3+ related samples (within-subject designs).

Question 8

Handling ties in ranks

Answer 8

-Assign the average rank across tied values
-Do not alter other ranks

This approach maintains the integrity of the ranking system.

Question 9

SPSS procedure non-parametric tests

Answer 9

-Analyse -> nonparametric tests -> choose appropriate test
-Ensure variables set to correct type (nominal, ordinal or scale)
-SPSS reports raw data and standardised stats

Ensuring you select the correct test is crucial for accurate results.