GANCQR: Estimating Prediction Intervals for Individual Treatment Effects with GANs

Evaluating individual treatment effects (ITE) is challenging due to the lack of access to counterfactual outcomes, particularly when working with biased data. Recent efforts have focused on leveraging the generative capabilities of models like Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs) for ITE estimation. However, few approaches effectively address the need for uncertainty quantification in these estimates. In this work, we introduce GANCQR, a GAN-based conformal prediction method that generates prediction intervals for ITE with reliable coverage. Numerical experiments on synthetic and semi-synthetic datasets demonstrate GANCQR's superiority in handling selection bias compared to state-of-the-art methods.