ABSTRACT

Chapter 3 introduces the potential outcomes framework for causal inference together with the Fundamental Problem of Causal Inference, which is that only one potential outcome, can possibly be observed per study participant. Causal concepts are presented and defined, including causal types, the randomization or stratified randomization assumptions, positivity, consistency, confounding by indication, conditional causal effect, causal effect measures such as the risk difference, relative risk, other relative risk, and odds ratio, number needed to treat, attributable fraction, probability of necessity, causal power, probability of sufficiency, probability of necessity and sufficiency, unconditional causal effect, measures of association, adjusted associations, the connection between linear models and the risk difference, the connection between loglinear models and the relative risk, the connection between logistic models and the odds ratio, and vaccine efficacy. Examples and R code are also provided.