Abstract: Causal inference is a core problem in empirical research in the social sciences,but understanding the context of causal inference relies on algebraic derivation,a fact that hinders the prevalence of causal knowledge among sociologists. Causal graphs derived from computer science can intuitively present casual paths and control strategies in a graphical way,thereby providing people with a non-parametric toolkit for understanding causal problems. This paper aims to provide a comprehensive introduction to the causal graph method and integrate it with the existing framework of causal inference based on regression models. This article first introduces the conceptual rules and the three basic configurations of chain,fork,and inverted fork that make up causal graphs. Then,it discusses the opening and blocking of causal pathways between variables and the three sources of bias that can mislead the identification of true causal relationships,namely confounding bias,over-control bias and endogenous selection bias. The article further introduces the D-separation rule used to determine which variables in a causal inference should be controlled. On this basis,various empirical examples are brought in to interpret four endogenous problems of omitting variable bias, sample selection bias,self-selection bias,and simultaneity bias through causal graphs. Graphic expressions and implementing conditions of several causal inference methods are also identified,including multiple regression and matching,proxy,experiments, instrumental variable,and panel models. In addition,this article attempts to clarify two common misconceptions in causal inference:conditioning on a post-treatment variable does not necessarily lead to bias and conditioning on a pre-treatment variable may cause deviation. Finally,it is suggested that the application of the causal graph method can help standardise causality research and facilitate the teaching and dissemination of causal inference knowledge.

Key words: causal graph, non-parametric causal inference, confounding bias, over-control bias, endogenous selection bias