ACIC 2018 Competition Datasets

Complete datasets from the 2018 Atlantic Causal Inference Conference (ACIC) Data Analysis Challenge, containing 2,592 simulated datasets from 64 Data Generating Processes with variable sample sizes (1,000–50,000 observations). Each dataset draws a subsample from a shared covariate matrix of 100,000 individuals and 177 features. Binary treatment assignment with continuous outcomes. Ground truth individual potential outcomes (y0, y1) are provided for each dataset. Archive structure: x.csv: shared covariate matrix (100,000 × 177) scaling/scaling/params.csv: DGP metadata and ground truth parameters for all 2,592 datasets scaling/scaling/factuals/: 2,592 CSV files with observed treatment and outcome (z, y) scaling/scaling/counterfactuals/: 2,592 CSV files with potential outcomes (y0, y1)