Causal Inference for Complex Data: Heterogeneity, Longitudinal Settings, and Beyond

Statistical modeling is ubiquitous across all areas of research, especially in the biomedical sciences, yet these analyses often carry the age-old warning that "correlation does not imply causation." In many real-world settings, understanding correlation is simply not enough. We need to understand whether X causes an effect on Y, but beyond clinical trials settings there are many factors we cannot control for that may muddle this relationship. Causal inference methods are designed for precisely this purpose. With new advances happening all the time, the causal inference methods themselves are highly visible in a variety of real-world settings. As they are applied broadly, new advances are needed to accommodate the usual data challenges like treatment effect heterogeneity, longitudinal and non-stationary data, and incorporation of external data in causal inference. Just as its goals are distinct from traditional inference about correlations, the ways in which these challenges are overcome in causal inference are unique. This session, sponsored by the Caucus for Women in Statistics, brings together women in our field to celebrate the new advances they are making in causal inference.