Observational-Interventional Priors for Dose-Response Learning
This work addresses the problem of efficient causal effect estimation for researchers in fields like healthcare, where controlled experiments are costly, by providing a method that leverages observational data to enhance learning from interventions, though it is incremental as it builds on existing nonparametric techniques.
The paper tackles the challenge of learning dose-response curves from limited controlled interventions and observational data by introducing a hierarchical Gaussian process prior that combines both sources, demonstrating improved learning speed through sensitivity analysis and an application to therapy effects on premature infants' cognitive skills.
Controlled interventions provide the most direct source of information for learning causal effects. In particular, a dose-response curve can be learned by varying the treatment level and observing the corresponding outcomes. However, interventions can be expensive and time-consuming. Observational data, where the treatment is not controlled by a known mechanism, is sometimes available. Under some strong assumptions, observational data allows for the estimation of dose-response curves. Estimating such curves nonparametrically is hard: sample sizes for controlled interventions may be small, while in the observational case a large number of measured confounders may need to be marginalized. In this paper, we introduce a hierarchical Gaussian process prior that constructs a distribution over the dose-response curve by learning from observational data, and reshapes the distribution with a nonparametric affine transform learned from controlled interventions. This function composition from different sources is shown to speed-up learning, which we demonstrate with a thorough sensitivity analysis and an application to modeling the effect of therapy on cognitive skills of premature infants.