A/B Testing and Causal Inference

10 Academy · 2022 · Academic project · Individual

A hypothesis-testing framework measuring ad-campaign brand lift, and a causal-inference study of delivery-fleet data using Bayesian networks and clustering.

A/B testing
Causal inference
Bayesian networks
Statistics

Overview

Two related analyses: a classical and sequential A/B testing framework for measuring the brand-awareness lift of ad campaigns, and a causal study of delivery-fleet location data asking which factors actually drive unfulfilled orders, using Bayesian networks to answer counterfactual questions.

Media

Causal graph of delivery variables — Learned Bayesian network over delivery features, used to ask counterfactual questions about unfulfilled orders.

Map of delivery request locations — Delivery request locations across the city, colored by volume.

Resources

A/B testing code on GitHubgithub
Causal inference code on GitHubgithub
Write-up: Causal Inference (Medium)article