#011 — Mann-Whitney U Test (Wilcoxon Rank-Sum)
Statistical Definition
Non-parametric test comparing the distributions (stochastic dominance) of two independent groups based on pooled rank order.
Bioinformatics Use Case
Single-cell RNA-seq differential expression testing of zero-inflated, skewed UMI counts between two cell clusters.
Input Variables
2 independent continuous or ordinal vectors (x1, x2) without normality requirement.
Application
Single-cell and clinical datasets are notoriously skewed and zero-inflated. The Mann-Whitney U test delivers rock-solid differential discovery without fragile normality assumptions.
#012 — Wilcoxon Signed-Rank Test
Statistical Definition
Non-parametric test evaluating whether the median difference between paired observations differs from zero using signed ranks.
Bioinformatics Use Case
Comparing matched tumor vs. adjacent normal tissue methylation beta-values across patients where differences are non-Gaussian.
Input Variables
2 matched numeric vectors (x1, x2) of equal length.
Application
When patient biopsy differences contain extreme outliers, the Wilcoxon Signed-Rank test preserves paired study power without letting a single outlier skew your p-value.
#013 — Kruskal-Wallis One-Way ANOVA by Ranks
Statistical Definition
Non-parametric alternative to One-Way ANOVA testing whether samples across three or more groups originate from the same distribution.
Bioinformatics Use Case
Comparing Shannon alpha-diversity indices of gut microbiomes across omnivore, vegetarian, and vegan cohorts.
Input Variables
1 numeric/ordinal response vector (y); 1 categorical grouping factor (k >= 3 levels).
Application
Microbiome diversity and clinical scores rarely follow bell curves. Kruskal-Wallis rigorously validates differences across three or more patient cohorts with zero distributional fragility.
#014 — Friedman Test
Statistical Definition
Non-parametric alternative to Repeated Measures ANOVA for comparing three or more matched treatments or timepoints across blocks.
Bioinformatics Use Case
Benchmarking F1-scores of 4 variant callers across 15 diverse whole-genome sequencing datasets.
Input Variables
Matrix of n blocks (rows) by k >= 3 treatments/algorithms (columns).
Application
When benchmarking competing algorithms or tracking non-normal longitudinal patient scores across time, the Friedman test provides bias-free, rank-based validation.
#015 — Kolmogorov-Smirnov Two-Sample Test
Statistical Definition
Non-parametric test quantifying the maximum vertical distance between the empirical cumulative distribution functions (ECDFs) of two samples.
Bioinformatics Use Case
Detecting shifts in insert-size distributions or cell-cycle DNA content histograms between treated and control cells.
Input Variables
2 independent continuous numeric vectors (x1, x2).
Application
Sometimes two treatments share the exact same average response while dramatically altering population tails or bimodality. The KS test catches any shift in distribution shape.
#016 — Anderson-Darling k-Sample Test
Statistical Definition
Tail-sensitive non-parametric test evaluating whether k samples are drawn from the same underlying continuous distribution.
Bioinformatics Use Case
Comparing extreme tail behavior of somatic mutation burden distributions across cancer types.
Input Variables
k >= 2 continuous numeric vectors.
Application
In oncology and rare-variant genomics, the biological signal lives in the extreme tails. Anderson-Darling weights tail deviations far more sensitively than standard KS tests.
#017 — Mood's Median Test
Statistical Definition
Non-parametric test checking whether two or more independent groups have identical medians by counting observations above and below the grand median.
Bioinformatics Use Case
Comparing median hospital length-of-stay or survival time across treatment arms in the presence of extreme infinite/censored outliers.
Input Variables
k >= 2 continuous or ordinal numeric vectors.
Application
When extreme outliers corrupt means and even distort rank sums, Mood's Median Test provides a clean, outlier-immune verdict on whether group medians truly differ.
#018 — Brunner-Munzel Test
Statistical Definition
Generalized Wilcoxon-Mann-Whitney test for stochastic equality that remains valid even when group variances and shapes differ.
Bioinformatics Use Case
Comparing skewed, heteroscedastic gene expression or FACS fluorescence intensities between two unequal cell populations.
Input Variables
2 independent continuous or ordinal vectors (x1, x2).
Application
Standard Mann-Whitney tests can misfire when two populations have different shapes or spreads. The Brunner-Munzel test guarantees accurate false-positive control under heteroscedastic skew
#019 — Jonckheere-Terpstra Trend Test [019_Jonckheere_Terpstra_Trend_Test.ipynb]
Statistical Definition
Non-parametric test specifically powered to detect a monotonic ordered trend across three or more sequential groups.
Bioinformatics Use Case
Testing for a dose-dependent increase in toxicity biomarker across Control, Low, Medium, and High drug concentrations.
Input Variables
k >= 3 ordered numeric vectors (g1, g2, ..., gk).
Application
Standard multi-group tests ignore dose order and waste statistical power. The Jonckheere-Terpstra test directly targets monotonic dose-response trends across escalating treatments.
#020 — Sign Test for Paired Medians [020_Sign_Test_for_Paired_Medians.ipynb]
Statistical Definition
Distribution-free binomial test evaluating whether the number of positive versus negative paired differences departs from 50/50.
Bioinformatics Use Case
Assessing directional consistency of pathway upregulation across heterogeneous patient-derived xenografts without assuming symmetry.
Input Variables
2 matched numeric vectors (x1, x2) of equal length.
Application
When paired differences are asymmetrically skewed and even signed-rank assumptions break down, the Sign Test gives an undeniable directional proof of treatment consis