Learning Joint Nonlinear Effects from Single-variable Interventions in the Presence of Hidden Confounders
This addresses causal inference challenges in fields like medicine or economics where hidden confounders are common, but it is incremental as it builds on existing intervention-based methods.
The paper tackles the problem of estimating effects of multiple simultaneous interventions with hidden confounders by using single-variable intervention data, proving identifiability under a nonlinear structural causal model with additive Gaussian noise and showing improved performance in experiments.
We propose an approach to estimate the effect of multiple simultaneous interventions in the presence of hidden confounders. To overcome the problem of hidden confounding, we consider the setting where we have access to not only the observational data but also sets of single-variable interventions in which each of the treatment variables is intervened on separately. We prove identifiability under the assumption that the data is generated from a nonlinear continuous structural causal model with additive Gaussian noise. In addition, we propose a simple parameter estimation method by pooling all the data from different regimes and jointly maximizing the combined likelihood. We also conduct comprehensive experiments to verify the identifiability result as well as to compare the performance of our approach against a baseline on both synthetic and real-world data.