查看文章

Our goal here is to motivate, explain, and offer an example of how social epidemiologists might use the propensity score approach. Methodologically, we offer nothing especially new or ground-breaking. Instead we aim to synthesize and make accessible existing research and show that the combined use of a counterfactual framework for causal inference, an explicit causal contrast study design, and propensity score matching methods is a useful alternative approach to regression models.“Accessibility” is essential, because the relevant literature is both vast and often quite technical if not impenetrable for non-statisticians. Even some of the published tutorials (for example, D’Agostino 1998; Joffe and Rosenbaum 1999; Little and Rubin 2000) can present challenges, and none are tailored for social epidemiologists. Because social epidemiology is clearly interested in estimating the effect of neighborhood contexts on health outcomes (that is, neighborhood effects), we use such an investigation as an example and departure point for discussion. Accordingly, we divide this chapter into four sections:(1) the counterfactual framework,(2) propensity score-matching methods,(3) example, and (4) conclusion.