This ability to emulate interventions by smart observations could not have been acquired had the statistical properties of Bayesian networks not been unveiled between 1980 and 1988. We can now decide which set of variables we must measure in order to predict the effects of interventions from observational studies.
