multiple-testing

Learn then Test: Calibrating Predictive Algorithms to Achieve Risk Control

The MultipleTesting package offers common algorithms for p-value adjustment and combination and more…

Solutions of applied exercises contained in "An Introduction to Statistical Learning with Applications in Python", by Tibshirani et al, edition 2023

Statistical inference for Sharpe ratios: probabilistic Sharpe ratio, minimum track record length, and FDR/FWER corrections for screening many strategies

Statistical inference of recent positive selection using IBD segments

Repository for R and Python packages and reproduction codes in Weighted Conformalized Selection paper

This repository contains a collection of functions to evaluate investment strategies regarding multiple testing concerns.

A Shiny app for graphical multiplicity control

Sequential Hypothesis Testing with e-Values and p-Values

A FDR controlling procedure based on hidden Markov random field (Biometrics-15 paper)

Causal Inference for Genomic Data with Multiple Heterogeneous Outcomes

Fixed Sequence Multiple Testing Procedures

Methods in my PhD thesis

Backtest overfitting audit for factor research: probability of backtest overfitting (PBO), deflated Sharpe ratio, point-in-time data, purged walk-forward. Searches published factor libraries and reports what actually survived costs.

Pre-registered falsification of the commodity carry premium (XS + TS) on 18 CME futures, 2010–2026 — design frozen before data; 0/2 arms promoted; four exchange-data identity traps documented along the way.

Most associations in this public semiconductor dataset are artifacts of time. 32 features clear BH-FDR with permutation p=0.0005, but only 2 survive a change of time period. Chronological split, DuckDB/SQL warehouse, SPC, cost-aware baseline, every transform fitted on train only.

A quantitative research platform built for falsification: a validation toolkit written from the papers (purged CPCV, PBO, deflated Sharpe, SPA, MCPT) and 23 pre-registered experiments, failures included.

POSSA: Power simulation for sequential analyses and multiple hypotheses.

NYU DS-GA 1020 Final Project