Much of epidemiology, econometrics, sociology and political science is an attempt to infer causal relationships using data gathered under conditions where fully controlled experiments are not possible ...