The study introduces a new method, the Conformal Monte Carlo (CMC) meta-learners, for estimating the treatment effect of interventions for decision-making. Traditional methods only provide a point estimate of this treatment effect, however, the CMC meta-learners leverage conformal predictive systems, Monte Carlo sampling, and CATE meta-learners to produce a predictive distribution usable in individualized decision-making. The CMC framework shows promising results, providing strong experimental coverage and small interval widths for true individual treatment effect estimates.
Publication date: 8 Feb 2024
Project Page: https://github.com/predict-idlab/cmc-learner
Paper: https://arxiv.org/pdf/2402.04906
