Definition: The Kruskal-Wallis test is a non-parametric statistical method used to determine if there are statistically significant differences between the medians of three or more independent groups on a continuous or ordinal dependent variable. It serves as an alternative to one-way ANOVA when the assumption of normality is violated or when dealing with ordinal data.
The Kruskal-Wallis test operates by ranking all observations from all groups together, from the smallest to the largest, regardless of group membership. It then calculates a test statistic (H-statistic) based on the sum of ranks for each group. If the group medians are similar, the average ranks for each group will be roughly equal; significant differences in average ranks suggest that at least one group median is distinct from the others. The null hypothesis states that the population medians of all groups are identical, while the alternative hypothesis suggests that at least one group’s median is different. This test is particularly valuable because it does not require assumptions about the distribution of the data (e.g., normality) or homogeneity of variances, making it a robust choice for analyzing skewed or non-normally distributed data often encountered in public health research.
In public health, the Kruskal-Wallis test is frequently employed when comparing health outcomes or risk factors across multiple independent groups where parametric assumptions cannot be met. For instance, it could be used to compare the median levels of a biomarker (e.g., inflammatory markers, pollutant exposure) among individuals from different socioeconomic status categories, or to assess differences in self-reported health scores (an ordinal variable) across various age groups or geographical regions. This test is also useful in evaluating the impact of multiple public health interventions, such as comparing the median improvement in physical activity levels after participating in three distinct health promotion programs. Its ability to handle non-normal and ordinal data makes it an indispensable tool for drawing reliable conclusions from diverse and often complex public health datasets.
Key Context:
- Non-parametric statistics
- One-way Analysis of Variance (ANOVA)
- Post-hoc tests (e.g., Dunn’s test)