Alexis Dinno's 'dunn.test' computes Dunn's test (1964) for stochastic dominance among multiple groups following a Kruskal-Wallis test. The method makes k(k-1)/2 pairwise comparisons using z-test approximations of rank statistics and accounts for tied ranks. Its null hypothesis corresponds to that of the Wilcoxon-Mann-Whitney rank-sum test.
Use Cases
- Conduct post-hoc analysis to identify specific group differences after a significant Kruskal-Wallis test result.
- Perform non-parametric multiple pairwise comparisons for data that does not meet parametric test assumptions.
- Test for median or mean differences between groups when distributions are identical except for location.
- Analyze ranked data with tied observations, as the method accounts for tied ranks.
Strengths
- Implements a well-established statistical method (Dunn's test, 1964) for rigorous multiple comparisons.
- Accounts for tied ranks, which is a common issue in non-parametric rank-based analysis.
- The description provides a clear statistical foundation, referencing the original works by Dunn, Kruskal, and Wallis.
Limitations
- Description metadata is limited; actual data quality requires manual inspection after download.
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
Provenance
- Source
- Alexis Dinno
- Collection Method
- Likely a software implementation or statistical package for performing Dunn's test.
- Time Range
- Methodology references years 1952 (Kruskal-Wallis) and 1964 (Dunn's test).
- Freshness
- Last update date is unknown; freshness unverified.
- Geography
- null