Introducing Causal Inference Using Bayesian Networks and <i>do-</i>Calculus
Yonggang Lu, Qiujie Zheng, Daniel M. Quinn
Abstract
We present an instructional approach to teaching causal inference using Bayesian networks and do-Calculus, which requires less prerequisite knowledge of statistics than existing approaches and can be consistently implemented in beginner to advanced levels courses. Moreover, this approach aims to address the central question in causal inference with an emphasis on probabilistic reasoning and causal assumption. It also reveals the relevance and distinction between causal and statistical inference. Using a freeware tool, we demonstrate our approach with five examples that instructors can use to introduce students at different levels to the conception of causality, motivate them to learn more concepts for causal inference, and demonstrate practical applications of causal inference. We also provide detailed suggestions on using the five examples in the classroom.