189040 Role of Bayes in more realistic inferences about exposure-disease relationships

Tuesday, October 28, 2008: 4:50 PM

Paul Gustafson, PhD , Department of Statistics, University of British Columbia, Vancouver, BC, Canada
One task that Bayesian methodology can accomplish in the epidemiologic context of inferring exposure-disease relationships is the replacement of more dubious assumptions with less dubious assumptions. That is, a prior distribution can be constructed to state than a standard assumption is not likely to be strongly violated, rather than stating that the assumption holds exactly. For instance, one can replace an assumption of no unmeasured confounders with an assumption that there are not likely to be strong confounders which aren't measured. One can replace an assumption that exposure assessment is perfect with an assumption that exposure assessment is likely very good. Or one can replace an assumption that exposure is misclassified with known misclassification probabilities with an assumption that the supplied values are good guesses for the misclassification probabilities. This talk will survey the pros and cons of employing Bayesian methods in this way.

Learning Objectives:
1. Recognize the basic structure of Bayesian statistical inference. 2. Recognize how Bayesian methods can be appropriate and useful in epidemiologic contexts.

Keywords: Statistics, Epidemiology

Presenting author's disclosure statement:

Qualified on the content I am responsible for because: Bayesian inference is my main area of research.
Any relevant financial relationships? No

I agree to comply with the American Public Health Association Conflict of Interest and Commercial Support Guidelines, and to disclose to the participants any off-label or experimental uses of a commercial product or service discussed in my presentation.