HSS 2381 Lecture Notes - Lecture 8: Continuity Correction, Grou, Null Hypothesis
Document Summary
Nonparametric tests: primarily used when 1) outcomes are not measure on an interval or ratio scale or: assumptions for parametric test are severely violated. Parametric data: compute a statistic from a sample and related it toa parameter in a reference population; assumes the sample comes from a known distribution. Nonparametric data: none of the above; ranked correlations, chi square, etc. Choosing which one to use depends on the level of measurement. Selection of a nonparametric test depends mostly on: number of groups being compared; 2 vs. 3 or more groups, type of comparison being made. Dependent groups (within-subjects/repeated measures, correlated groups designs: tests for independent groups. Two groups + nominal measures = chi-square test. Two groups + ordinal measure = mann-whitney u test. Three or more groups + nominal measures = chi-square test. Three or more groups + ordinal measures = kruskal-wallis test.