Non-Parametric Approaches and Test Selection

Non-parametric analyses rely on many rules and assumptions.

Parametric analyses are preferred when possible.

Independence of observations refers to measurements influencing each other.

Which of the following is not a requirement for parametric analyses?

Factorial ANOVAs are generally robust to minor deviations from normality.

What is the appropriate alternative test for a one-way ANOVA if the assumptions are not met?

Which test should you use for comparing the mean difference scores of a repeated measures design when the assumptions of normality are not met?

Non-parametric tests use the median to account for extreme values that can skew the mean.

What test is used to check for homogeneity of variance?

Which test is appropriate for comparing the medians of two independent groups?

Which test is appropriate for comparing the medians of three or more independent groups?

Which test is appropriate for comparing the medians of two related groups, such as pre-test and post-test scores?

Which test is appropriate for comparing the medians of three or more related groups, such as measuring performance at multiple time points?

Non-Parametric Approaches

Non-parametric analyses do not rely on many assumptions, making them less complex
They are less useful for intricate study designs.
Parametric analyses are generally preferred when possible
Parametric analyses assume independence of observations (measurements do not influence each other).
Parametric analyses require interval or ratio level data, which are measured on a scale with meaningful intervals.
Non-parametric analyses are useful when data is categorical or ordinal, using ranks instead of actual scores.

Choosing the Right Test

Factorial ANOVAs are generally robust to deviations from normality.
For one-way ANOVAs and t-tests, alternative non-parametric tests can be used to handle skewed data.
The t-test looks at normally distributed independent data. Repeated measures t-tests consider the differences.
Example of research questions that might use non-parametric approaches: How many hours do students spend revising for psychology, geography or biology exams.

Non-Parametric Tests for Specific Designs

Kruskal-Wallis, Wilcoxon, Mann-Whitney are non-parametric tests for various designs, as shown in the flow chart.
These tests are utilized when the assumptions of parametric tests are not met.
The chart guides researchers to choose the appropriate tests based on the design (e.g., 2 conditions vs. 3+ conditions; independent or repeated measures).
The appropriate tests are recommended after calculating the median and using appropriate post-hoc analysis tests (Mann-Whitney for Kruskal-Wallis, Wilcoxon Signed-Rank).

This quiz covers non-parametric approaches in statistical analysis, highlighting their assumptions and comparison to parametric methods. It explores when to choose non-parametric tests over traditional tests like ANOVA and t-tests, especially in handling different data types. Enhance your understanding of statistical methodologies with practical examples.