propensity-score-matching

Propensity Score Matching(PSM) on python

Predictive State Propensity Subclassification (PSPS): A causal deep learning algoritm in TensorFlow keras

This repository commits to the application of biostatistics knowledge on clinical, randomized trials and observational studies.

Lecture slides, video recordings, and coding exercises from the 2024 Northwestern University Causal Inference Workshop. This repository is not affiliated with Northwestern University or the workshop.

📊 EconKit — 一站式计量经济学实证分析工具 | All-in-one Econometrics Toolkit for Chinese Economics Students. DID/PSM/RDD/IV/FE/GMM, parallel trends, placebo tests, PDF report. No code needed.

R biostatistics project: propensity score matching + Cox proportional hazards survival analysis on a SQL-extracted synthetic claims cohort, grounded in clinical/epidemiological literature review.

Review of propensity score matching techniques, especially IPTW, Overlap Weights, and Doubly Robust Estimation

Staggered-adoption causal platform: TWFE DiD, event study, parallel-trends and not-yet-treated ATTs + PSM in Python (statsmodels); TypeScript/Next.js dashboard with Zod validation; Jest + Playwright + pytest in Docker CI. Headline: GMV +6.3-8.2% (95% CI) on panel data.

Does switching to rail actually cut down how much someone flies, or is that just a nice story correlation tells us? Built to find out properly, with propensity score matching, a validation check against a known answer, and honesty about what it still can't prove.

This repository is associated with propensity score analysis utilizing machine learning algorithms to examine the impact of health insurance on duration of hospital stay in the NYC SPARCS 2015 In-patient discharges dataset.

Causal inference project using DoWhy to isolate the true marketing lift of bank contact methods. Applies Propensity Score Stratification to remove selection bias from raw campaign data and delivers an interactive ROI simulator for budget decision-making.

Code and supporting files used as part of the impact evaluation of the Detect, Protect and Perfect (DPP) project to reduce stroke incidence by improving care of atrial fibrillation.

HSE FES Econometrics-2 (advanced course) project

Statistical Evaluation Methods

This repository analyzes a marketing campaign using Propensity Score Matching, heterogeneous treatment effects, and causal forests to estimate the true impact of an algorithm across customer segments and channels. The project converts raw transactional and engagement data into actionable deployment recommendations and revenue impact estimates.

Propensity score matching estimators for QTE, QTT, ATE, and ATT in R.

Command-line guardrail for online experiments. Checks traffic allocation for sample ratio mismatch, runs the metric test that suits each metric's shape, corrects for multiple testing, re-estimates the effect under propensity score matching, and reports one of three verdicts. Validated against the 13.98M-row Criteo Uplift dataset.

New-user active-comparator cohort study comparing SGLT2i vs DPP4i on HF hospitalization risk. Propensity score matching, Cox PH, IPTW — full protocol, SAP, and results report in R.

营销渠道因果增量分析：朴素相关 vs 回归调整/PSM/IPW，UCI银行营销4.1万条真实数据。Causal uplift of marketing channel — confounding-adjusted (PSM/IPW) on real data.

A comparative performance study of Propensity Score Matching, Doubly Robust Estimation, and Stratification algorithms. Evaluates Average Treatment Effect (ATE) accuracy and computational runtime across high-dimensional and low-dimensional datasets using L1-penalized propensity score estimation.