Thursday, August 20 2009, 3:30pm Dobbin Epidemiology and Biostatistics, University of Georgia There are now several methods for constructing confidence intervals for prediction accuracy in high dimensional settings. But these methods have high computational cost and are cumbersome to implement. As a result, these types of intervals are rarely reported, and their properties are not well understood. In this talk, we review these methods, one in some detail, and introduce current work which utilizes a mathematical modeling approach to try to reduce the computational cost.
