Causal inference in case-control studies

摘要: We investigate partial identification of causal relative and attributable risk---the ratio two counterfactual proportions the difference between them---in case-control case-population studies. The odds is shown to be a sharp upper bound on risk under monotone treatment response selection assumptions, without resorting strong ignorability, nor rare-disease assumption. Sharp bounds are also obtained same assumptions. Paying special attention (conditional) ratio, we propose semiparametrically efficient estimator aggregated (log) ratio. Further, develop easy-to-implement inference procedures for risk. Finally, showcase our methodology by applying it unique datasets in literature. find that attending private school may have little effect entering very selective university Pakistan dropping out could substantially increase joining criminal gang Brazil.