Bias Reduction for High-Dimensional Multivariate Data

• Activity Number: 682
• Type: Contributed
• Date/Time: Thursday, August 5, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Graphics
• Author(s): Siamak Noorbaloochi*+ and David Nelson
• Companies: VA Medical Center, Minneapolis and VA Medical Center, Minneapolis
• Keywords: High Dimensional ; Selection bias ; Multivariate distributions

Abstract

Directional bias for multivariate data occurs when the sample selection mechanism is a monotone function of each of its components. Assuming directional bias and a high-dimensional multivariate model for the unbiased population, we first propose an information-preserving methodology to replace the original model with a univariate distribution and then, we extend size-bias theory to enable making inference about the underlying population when a biased and possibly an unbiased sample are available. Specifically, we will develop some estimation and testing methods for estimating the original unbiased population parameters. We will use simulated and real data to illustrate the procedures.