An Aligned Rank Transform Procedure for Multifactor Contrast Tests
summary
Introduces ART-C a new aligned-rank-contrast algorithm that claims to let aligned rank transform handle post-hoc contrasts across multiple factors without inflating false positives; validated against common tests over 72,000 synthetic data sets varying distributions, layouts, sample sizes, and within/between designs; finds does not inflate Type I error (except with Cauchy data) and higher power than t, Mann–Whitney U, Wilcoxon and ART; ships as code and as an extension to ARTool (R and Windows).
AI-written summary · about
related talks
45mTools for ART - Autonomous Real-time Testing
26mDiscovering Unwarranted Associations in Data-Driven...
18mHow Data Scientists Broke A/B Testing (And How We Can Fix It)
09m“Artful Testing”
28mCorrelation Sucks - Long Live Causality! Mastering Causal Inference in Data Science
24mOffline Ranking Validation - Predicting A/B Test Results
27mHow to A/B test with confidence
48mControlling FDR While Highlighting Selected Discoveries