ABSTRACT
This chapter describes the methods of assessing the exposure-outcome relationship in a multivariable fashion controlling for potential confounding. Using the appropriate statistical techniques depending on whether the potential confounders are continuous or categorical, a simple hypothesis test can assess whether there is a statistical relationship. From the birthwt dataset, confounder selection will model the relationship with the exposure of early pregnancy smoking and the outcome of low birth weight. Logistic regression is appropriate to model the cumulative incidence if there is complete follow-up for all study participants, with no withdrawals, and the time to event is unimportant. In a case-control study, the odds of exposure are assessed given the dichotomous outcome of being a case or a control. Although incidence is not measured in a case-control study, the goal of epidemiological inference is that the risk factor is causally related to the outcome, and therefore corresponds to an implicit notion of inference.
