Milo Schield
Session Ended
Location:
Information
Most students are more interested in causal inference than in population inference: inferring population statistics based on sample statistics. We will review three approaches: Rubin's potential outcomes+propensity scores, Pearl's DAGs and Do-calculus, and Pearson OLS. We will examine the accessibility of each. We evaluate whether the Wainer-Lesser graph is the best way to show students how a statistical two-group comparison can be influenced after taking into account (controlling for) a binary confounder. We will discuss the need for stat-ed to offer a new course in causal inference.