摘要
本文旨在介紹因果推論的核心思維及其在生醫方面的應用。第一章從日常對話中對於「因果」的直覺開始，引導讀者去思考因果關係這個概念需要哪些條件才能成立、或著說我們究竟觀察到什麼樣的情況會認可某種關聯的確是因果關係。第二章我們正式介紹因果推論中會用到的術語，像是「介入」或是「潛在結果」等概念；並使用實例讓讀者理解作因果推論所必須的兩大假設，也就是「可交換性」跟「一致性」。第三章我們進入本文的主題，也就是存在於「因果之間」的中介因子如何協助我們拆解目標機制，也就是介紹「因果中介模型」。我們講解了從基礎因果模型延伸到因果中介模型所需要的廣義版可交換性假設，以及介紹了更加廣義的「多重中介因子模型」，與其相關的估計及檢定問題。最後，我們使用實驗室新發表的一篇討論新冠疫苗、抗體量與疫苗保護力的文章做為範例，來介紹探討因果中介的研究如何進行，並稍加簡介從資料挖掘出因果關係的「因果發現」領域。希望這篇文章能深入淺出的引領讀者對因果推論產生興趣。
並列摘要
This article aims to introduce the essence of causal inference and its applications in the biomedical field to the general public. Section 1 begins by exploring the intuitive understanding of "causality" in everyday conversations. We guide the readers to contemplate the conditions required for establishing a causal relationship and the scenario that justifies recognizing a certain correlation as a causal relationship. Section 2 formally introduces the terminology used in causal inference, such as "intervention" and "potential outcomes," while utilizing examples to help readers grasp the concepts of two fundamental assumptions required for causal inference: "exchangeability" and "consistency." Section 3 delves into the main theme of this article, which is how mediators "dwelling between causes and effects" can assist in dissecting the underlying mechanisms. We introduce the "causal mediation model" and discuss the generalized version of exchangeability required to extend from basic causal models to causal mediation models. Additionally, we also cover the estimation and testing issues related to the general "multimediator models." Finally, we present our recently published article discussing the relationship among the COVID-19 vaccine, antibody levels, and vaccine efficacy to illustrate how causal mediation analyses are conducted. We briefly introduce the field of "causal discovery," which uncovers causal relationships from data mining. Our hope is that this article provides an in-depth and engaging introduction, sparking readers' interest in causal inference.