ABSTRACT
Inference for causal effects is extremely important in the clinical sciences and in public health. The field of statistics is critical for addressing such problems using data from both randomized experiments and observational studies. Well-done randomized experiments provide the gold standard for inferring causal effects, whereas drawing such inferences, even in ideal observational studies, requires great care. Here we use the widely utilized Rubin Causal Model (RCM) to define causal effects in both randomized experiments and observational studies, and to draw causal inferences in both settings. The required statistical techniques, although conceptually analogous, can dramatically differ in practice between randomized experiments and observational studies. However, we use well-accepted principles of design and analysis in experiments to bridge to the design and analysis of observational studies, which is the context of many causal inferences in clinical studies.
