We thank all participants!
The Fourth Annual Atlantic Causal Inference Conference (ACIC) Data Challenge provides an opportunity to compare causal inference methodologies across a variety of data generating processes (DGP). As in previous years, the challenge focuses on computational methods of inferring causal effects from quasi-real world data. This year the target of estimation is the population average treatment effect (ATE).
Covariates were drawn from publicly available data and also simulated. Identifiability of the parameter is guaranteed, however challenges to estimation have been built-in to the processes for generating the binary treatment assignment, and binary or continuous outcome. These include non-linearity of the response surface, treatment effect heterogeneity, varying proportion of true confounders among the observed covariates, and near violations of the positivity assumption.
This year's challenge has two tracks:
Participants (individuals or teams) will download the datasets for either track (3200 datasets per track), and run analyses using their own computing resources to estimate the ATE and a 95% confidence interval for each dataset. Results for all datasets within a single track should be saved to a file, then uploaded to the website for evaluation.
Timeline
The 2019 Data Challenge is open mid-January through mid-April. Preliminary results will be announced at ACIC, 2019. The conference will take place at the McGill University, May 22 to May 24, 2019.
Key Dates
Organizing Committee:
Contact us: sgruber[at]putnamds[dot]com