About the speaker
Stephen R. Cole is a Professor in Epidemiology at UNC Gillings School of Global Public Health. Stephen R. Cole works to build accurate and impactful knowledge, particularly population-health (epidemiologic) knowledge. Professor Cole is interested in study designs and analyses that accurately estimate parameters of central interest to population-health scientists, such as risk. These study designs include randomized experiments, pseudoexperiments (i.e., observational studies) and thought-experiments (e.g., simulation studies).
About the talk
What is a good question? Provisionally, good questions have proper syntax, are semantically meaningful, answerable, and important. Good epidemiologic questions require a measurable outcome, including the context; treatments or actions, if causal; and translate unambiguously to a parameter of interest. What is a good answer? Again provisionally, good answers are as robust and as accurate as possible. We present results from a simulation experiment of a minimal nontrivial example, motivated by epidemiology, which illustrates some features of good answers using estimators from nonparametric, parametric, and semiparametric models. In summary, an estimator from a semiparametric model performed best among candidate estimators that were not omnisciently correct.