Fallacies of Selection: Cases studies in biostatistics
Abstract
Selection bias arises when the effects of selection of variables or models on subsequent statistical analyses are ignored, i.e., failure to take into account "double dipping" of the data when assessing statistical evidence. Eighty years ago, the prominent statistician and mathematical economist Harold Hotelling (while at Columbia) drew attention to this issue. In recent year, there has been a concerted effort to address the problem, giving rise to the nascent field of post-selection inference. I will review some of my work in this area, with particular attention to biostatistical problems involving large-scale case-control studies and the discovery of networks of co-expressed and co-regulated genes.
Biography
Prof. lan W. McKeague's research interests include post-selection inference, empirical likelihood, order-restricted inference, non-standard asymptotics, statistical methods in physical oceanography, functional data analysis, inference for stochastic processes, survival analysis, competing risks models for HIV/AIDS data, Markov chain Monte Carlo and Bayesian methods, efficient estimation for semiparametric models, missing data, counting processes and spatial point processes. He is a Fellow of the Institute of Mathematical Statistics and a Fellow of the American Statistical Association.