多元回归分析的魅力在于:通过保持其他干扰因素不变,让社会科学的研究者在非实验情景之下,完成了自然科学家在实验室里才能做到的事情。——Wooldridge,2009正如Wooldridge所言,当我们试图利用观察数据(observational data)进行因果推论时,统计控制扮演着类似于实验控制的角色。从本质上来看,这两类控制方法的核心思想与Mill(1848)提出的“求异法”(methodof difference)一致。Mill告诉我们,在 The power of multivariate regression analysis is that by keeping other disturbing factors constant, researchers in the social sciences can accomplish what scientists in the laboratory can do in a non-experimental setting. - Wooldridge, 2009 As Wooldridge puts it, when we try to make causal inferences using observational data, statistical control plays a role that resembles experimental control. In essence, the core ideas of these two types of control methods are consistent with Mill’s (1848) proposed “method of difference.” Mill tells us, at