Machine learning approaches to heterogeneous causal inference
Causal forests are designed to estimate HTE by adapting decision trees to focus on treatment effect variation across subpopulations. Unlike standard trees, causal forests split data to capture differences in treatment effects, estimating CATE for more targeted insights. By building multiple trees, causal forests identify how interventions impact different groups. Let’s start with a recap of HTE as it is related to causal forests.
HTE refers to the variation in treatment effects across different subpopulations or individuals. Unlike ATE, which provides a single estimate of the treatment effect across the entire population, HTE recognizes that the effect of an intervention can differ based on various characteristics or contexts. For example, a medication might be more effective for younger patients compared to older ones, or a marketing campaign might work better for a particular demographic group.
HTE is formally...
