At the end of this session, you should be able to: • Define nonparametric test • Identify the types of nonparametric tests and the assumptions involved • Understand how to use techniques to test for significant differences
analyze nominal or ordinal data and draw statistical conclusions
more efficient and powerful than the corresponding parametric test
require no assumptions about the population probability distributions
• if the normality assumption grossly violated
• distribution-free methods
based on differences in medians
provide a well-foundationed way to deal with circumstance • in which parametric methods perform poorly
4
Data interval level scaling
Smaller samples No stringent assumptions
Numbers of stringent assumptions With correct assumptions (e.g., normal distribution), • parametric methods will be more efficient/ powerful than non-parametric methods
Convert raw values to ranks and then analyse ranks
does not require any specific conditions concerning the shape of populations does not require value of any population parameters
Nonparametric
Parametric
Sign test
One samples t-test
Wilcoxon Signed Rank test
Paired t-test
MannWhitney U
Independent t-test
Kruskal Wallis
One-way ANOVA