Commission Statistical Analysis & Publication-Grade Reproducible Notebooks
We provide consulting expertise in statistical analysis. Click on the category links below to see examples of biological use cases and associated statistical tests in out portfolio. Every test is defined and illustrated with an use case, their input variables and application example.
1. Parametric Mean & Location Comparison (001–010)
One-Sample Student's t-Test — Tests whether a single sample mean (e.g., normalized log_2 fold-change) differs from a reference constant.
Two-Sample Independent Student's t-Test — Compares means between two independent groups assuming equal variance.
Welch's Two-Sample t-Test — Compares means between two independent cohorts with unequal variances (e.g., controls vs. heterogeneous tumors).
Paired Sample t-Test — Evaluates mean differences across matched pre-/post-treatment biopsies or paired tumor/normal tissues.
One-Way Analysis of Variance (ANOVA) — Compares means across three or more independent molecular subtypes or treatment arms.
Welch's One-Way ANOVA — Multi-group mean comparison robust to heteroscedasticity across disease severity stages.
Two-Way Factorial ANOVA — Isolates main effects and synthetic-lethal interactions across two categorical factors (e.g., Genotype times Drug).
Repeated Measures ANOVA — Tracks longitudinal within-subject trajectories across three or more timepoints.
Analysis of Covariance (ANCOVA) — Compares group means while adjusting for continuous clinical confounders (e.g., age, baseline tumor volume).
Multivariate Analysis of Variance (MANOVA) — Simultaneously tests group mean vectors across multiple correlated metabolites or cytokines.
2. Non-Parametric & Rank-Based Tests (011–020)
Mann-Whitney U Test (Wilcoxon Rank-Sum) — Compares distributions of two independent skewed or zero-inflated single-cell/clinical cohorts.
Wilcoxon Signed-Rank Test — Evaluates median paired differences in non-Gaussian methylation or expression data.
Kruskal-Wallis One-Way ANOVA by Ranks — Non-parametric comparison across three or more cohorts (e.g., microbiome alpha-diversity).
Friedman Test — Non-parametric comparison across three or more matched treatments or benchmarked variant callers.
Kolmogorov-Smirnov Two-Sample Test — Detects any shift in empirical cumulative distribution shape, spread, or bimodality between two cohorts.
Anderson-Darling k-Sample Test — Tail-weighted distribution comparison sensitive to extreme somatic mutation burden outliers.
Mood's Median Test — Outlier-immune comparison of group medians via grand-median contingency counts.
Brunner-Munzel Test — Generalized rank test valid under unequal variances and differing distribution shapes.
Jonckheere-Terpstra Trend Test — Tests for monotonic ordered trends across escalating drug concentrations or disease stages.
Sign Test for Paired Medians — Distribution-free binomial test of directional consistency across asymmetrically skewed paired differences.
3. Categorical, Contingency & Enrichment Tests (021–032)
Pearson's Chi-Square Test of Independence — Evaluates association between categorical variables (e.g., mutation status vs. tumor grade).
Fisher's Exact Test — Exact hypergeometric test for 2 times 2 tables in small cohorts and Gene Ontology over-representation.
Hypergeometric Enrichment Test — Computes exact pathway over-representation probabilities for differentially expressed gene lists.
McNemar's Paired Contingency Test — Compares discordant binary classifications between two diagnostic assays on the same patients.
Cochran-Mantel-Haenszel (CMH) Test — Tests consistent binary association across stratified multi-center or multi-ancestry 2-times-2 tables.
Cochran-Armitage Trend Test — Tests linear additive genetic risk across ordered allele dosages (AA, Aa, aa) in GWAS.
G-Test (Log-Likelihood Ratio Contingency Test) — Information-theoretic contingency test for sequence codon usage and nucleotide composition.
Exact Binomial Proportion Test — Tests whether observed counts (e.g., allele-specific RNA-seq reads) deviate from a hypothesized proportion.
Two-Sample Z-Test for Proportions — Compares clinical responder rates between treatment and control arms.
Hardy-Weinberg Equilibrium (HWE) Exact Test — Quality-control test detecting genotyping artifacts in diploid SNP calls.
Barnard's Unconditional Exact Test — Higher-power unconditional alternative to Fisher's Exact Test for small 2 times 2 clinical trials.
Cochran's Q Test for Related Proportions — Compares binary detection sensitivity across three or more pipelines on the same truth set.
4. Correlation & Association Tests (033–042)
Pearson Linear Correlation Test — Quantifies linear coupling between continuous variables (e.g., promoter methylation vs. mRNA expression).
Spearman Rank Correlation Test — Measures monotonic association across non-linear or saturated biological dose-response curves.
Kendall's Tau Rank Correlation Test — Concordance-based rank correlation robust to ties in CRISPR screen rankings or clinical staging.
Partial Correlation Test — Isolates direct gene-gene co-expression after regressing out tumor purity or RIN batch confounders.
Point-Biserial Correlation Test — Quantifies association between a binary mutation status and a continuous drug IC_50 readout.
Distance Correlation (dCor) Permutation Test — Detects non-linear and non-monotonic (e.g., U-shaped or oscillatory) molecular dependencies.
Mantel Matrix Correlation Test — Permutation test correlating pairwise microbiome dissimilarity matrices against host genetic or spatial distances.
Canonical Correlation Analysis (CCA) Wilks' Lambda Test — Identifies and tests maximally correlated multi-analyte signatures across two omics layers.
Fisher's Z-Test for Comparing Two Correlations — Tests differential co-expression (loss or gain of regulatory coupling) between tumor and normal cohorts.
Intraclass Correlation Coefficient (ICC) Test — Evaluates technical assay reproducibility across replicates, plates, or laboratory sites.
5. Regression & Generalized Linear Models (043–054)
Ordinary Least Squares (OLS) Linear Regression t/F-Test — Models continuous expression or drug response against copy-number dosage or covariates.
Logistic Regression Wald & Likelihood Ratio Test — Predicts binary clinical outcomes or treatment response from multi-gene biomarker panels.
Poisson Regression Rate Ratio Test — Models somatic mutation counts per megabase adjusted for sequencing coverage offsets.
Negative Binomial GLM (DESeq2 / edgeR Core Test) — Models overdispersed bulk RNA-seq, ATAC-seq, and ChIP-seq read counts.
Zero-Inflated Negative Binomial (ZINB) Test — Separates technical dropout zeros from true abundance shifts in scRNA-seq and microbiome counts.
Beta Regression Significance Test — Models continuous proportions bounded in (0, 1) such as DNA methylation beta-values and cell fractions.
Linear Mixed-Effects Model (LMM) Wald / LRT — Accounts for donor-level random effects in multi-donor single-cell and hierarchical studies.
Generalized Estimating Equations (GEE) Wald Test — Estimates population-averaged marginal effects in longitudinal cohorts with correlated binary readouts.
Quantile Regression Significance Test — Models upper-tail percentiles (e.g., 90th percentile viral load or mutation burden) robust to outliers.
Nested Model Likelihood Ratio Test (LRT) — Proves whether a novel biomarker panel adds predictive power beyond standard clinical staging.
Firth Penalized Logistic Regression Test — Resolves infinite parameter estimates under complete separation in rare-variant case-control studies.
Generalized Additive Model (GAM) Smooth Term Test — Tests non-linear gene expression trajectories along single-cell pseudotime (tradeSeq).
6. Survival & Time-to-Event Analysis (055–062)
Log-Rank (Mantel-Cox) Survival Test — Compares right-censored Kaplan-Meier survival curves across biomarker-stratified patient groups.
Cox Proportional Hazards Wald & Score Test — Estimates multivariable Hazard Ratios adjusted for patient age, sex, and tumor stage.
Gehan-Breslow-Wilcoxon Survival Test — Risk-set-weighted survival test sensitive to early clinical separation in acute or immuno-oncology trials.
Tarone-Ware Survival Test — Square-root risk-weighted survival test balancing early and late event sensitivity.
Peto-Peto Prentice Survival Test — Survival-weighted test resistant to unequal patient censoring rates across treatment arms.
Schoenfeld Residuals Proportional Hazards Test — Diagnostic test verifying that Cox Hazard Ratios remain constant over follow-up time.
Restricted Mean Survival Time (RMST) Difference Test — Model-free comparison of months of survival gained up to horizon tau, valid when curves cross.
Gray's Competing Risks Cumulative Incidence Test — Isolates disease-specific relapse incidence in the presence of competing non-relapse mortality.
7. Distribution, Normality & Variance Tests (063–070)
Shapiro-Wilk Normality Test — Tests Gaussianity of normalized qPCR or proteomics data prior to parametric testing.
D'Agostino's K-Squared Omnibus Normality Test — Jointly tests skewness and kurtosis departures in large quantitative trait cohorts.
Levene's Test for Homogeneity of Variance — Detects differential transcriptional noise/dispersion across aging or disease states.
Bartlett's Test for Equality of Variances — Parametric test of variance stability across mass-spectrometry runs or sequencing batches.
F-Test for Equality of Two Variances — Compares technical measurement precision between two diagnostic assays.
Fligner-Killeen Non-Parametric Variance Test — Rank-based variance homogeneity test immune to heavy skew in single-cell data.
Hartigans' Dip Test for Unimodality — Detects bimodal ON/OFF transcriptional switches or hidden patient subpopulations.
Chi-Square Goodness-of-Fit Test — Evaluates whether observed variant or genotype frequencies match theoretical ratios.
8. Genomics, Transcriptomics & Sequence-Specific Tests (071–082)
Moderated Empirical Bayes t-Test (limma Core) — Shrinks gene-wise variances across all features to stabilize small-n omics inference.
Gene Set Enrichment Analysis (GSEA) KS Permutation Test — Detects coordinated pathway enrichment across ranked whole-transcriptome lists.
Permutational Multivariate ANOVA (PERMANOVA / adonis) — Tests beta-diversity centroid shifts across microbiome or single-cell distance matrices.
ANOSIM (Analysis of Similarities) — Rank-based permutation test quantifying between-group versus within-group community separation.
Tajima's D Neutrality Test — Scans population sequencing alignments for selective sweeps or balancing selection.
McDonald-Kreitman (MK) Selection Test — Compares within-species polymorphism to between-species divergence to detect adaptive protein evolution.
dN/dS (Ka/Ks) Codon Substitution LRT — Tests whether nonsynonymous-to-synonymous substitution rates deviate from neutral evolution (omega = 1).
Sequence Kernel Association Test (SKAT) — Variance-component score test aggregating bidirectional rare variants across genes in WES/WGS.
Transmission Disequilibrium Test (TDT) — Family-based trio association test immune to population stratification.
LD Score Regression (LDSC) Genetic Correlation Test — Separates polygenic heritability from confounding using GWAS summary statistics.
Colocalization Posterior Probability Test (coloc) — Bayesian test evaluating whether a GWAS locus and an eQTL share a single causal variant.
Mendelian Randomization Inverse-Variance Weighted (IVW) Test — Uses genetic instruments to infer causal effects of biomarkers on clinical disease.
9. Diagnostic Classification, ROC & Model Evaluation (083–088)
DeLong's Test for Correlated ROC AUC Curves — Head-to-head comparison of diagnostic AUROCs evaluated on the exact same patient cohort.
Hosmer-Lemeshow Calibration Goodness-of-Fit Test — Verifies that predicted clinical risk probabilities match observed event frequencies across deciles.
Cohen's Kappa Inter-Rater Agreement Test — Quantifies concordance between AI classifiers and pathologists beyond chance agreement.
Fleiss' Kappa Multi-Rater Agreement Test — Evaluates consensus subtype classification across three or more raters or clustering methods.
Brier Score & Calibration Slope Permutation Test — Assesses probabilistic prediction accuracy and calibration improvement over null baselines.
Matthews Correlation Coefficient (MCC) Significance Test — Balanced classification evaluation under severe class imbalance via chi square equivalence.
10. Time-Series, Spatial, Multi-Testing & Resampling Tests (089–100)
Granger Causality F-Test — Infers directed regulatory cascades from lagged longitudinal gene expression or metabolomics time series.
Augmented Dickey-Fuller (ADF) Stationarity Test — Tests for unit-root drift in longitudinal physiological telemetry or microbiome series.
Ljung-Box Autocorrelation Portmanteau Test — Diagnoses residual temporal autocorrelation in circadian or pharmacokinetic models.
Moran's I Spatial Autocorrelation Test — Identifies spatially variable genes and tissue niches in spatial transcriptomics (Visium/MERFISH).
Ripley's K / L-Function Spatial Point Pattern Test — Quantifies spatial clustering or exclusion of immune cells in multiplex immunofluorescence.
Benjamini-Hochberg False Discovery Rate (FDR) Procedure — Controls expected false discovery proportions across tens of thousands of omics features.
Benjamini-Yekutieli Dependent FDR Procedure — Controls FDR under arbitrary negative or complex feature correlations (e.g., compositional microbiome).
Holm-Bonferroni Step-Down FWER Test — Strictly controls family-wise error rate with higher power than standard Bonferroni.
Tukey's Honestly Significant Difference (HSD) Post-Hoc Test — Identifies specific differing group pairs following a significant multi-arm ANOVA.
Dunn's Non-Parametric Post-Hoc Multiple Comparison Test — Pinpoints pairwise rank differences following a significant Kruskal-Wallis test.
Exact / Monte Carlo Permutation Test of Location — Builds empirical null distributions for custom bioinformatics metrics via label shuffling.
BCa Bootstrap Confidence Interval & Hypothesis Test — Bias-corrected and accelerated resampling for uncertainty around complex biomarkers (IC50, Shannon entropy, AUROC).