Why Bayesian Statistics?
When I came to Duke as an assistant professor in 1992, I began collaborating with researchers trained in medicine and the social sciences. In these collaborations I was struck by one thing in particular: While the statistical analyses they presented in publications were nearly 100% classical, the statistical interpretations they made in their day-to-day work were not. In daily conversations, debates, and statistical analyses, they rarely followed classical prescriptions for ‘legitimate’ data analyses or gave classical interpretations to their inference. In their day-to-day activities, their thinking and the decisions they made based on this thinking were nearly 100% Bayesian.
Statistics are a tool for more than hypothesis testing; they are a tool for decision making. Bayesian inference provides a framework for decision making that is informed by prior knowledge and that is adjusted as additional evidence is gathered – exactly the way that we naturally make decisions. My intellectual interest is in promoting Bayesian statistical methods through undergraduate education and statistical practice.
In my research, I promote and facilitate use of Bayesian methods in disciplines including mental health, health policy, and the social sciences. Here are two books I have edited on the subject.