Peter Buhlmann - Heterogeneity: New opportunities for causal inference and prediction

Perhaps unexpectedly, heterogeneity in potentially large-scale data can be beneficially exploited for causal inference and more robust prediction. The key idea relies on invariance and stability across different heterogeneous regimes or sub-populations. What we term as "anchor regression" opens some novel insights and connections between causality and protection (robustness) against worst case interventions. The resulting new procedures offer (possibly conservative) confidence guarantees. We will discuss the novel methodology as well as some applications.