Robust Estimation"
Abstract
In this lecture we look at robust regression methods to automatically account for potential outliers.
Readings: Readings: Christensen Chapter 13 and Hoff Chapter 9
In this lecture, we look at model averaging for accounting for uncertainty about outliers represented as mean shift outliers. By placing a prior on the unknown mean, we show that mean shift outliers are equivalent to variance inflation outliers. We extend this model to allow continuous variance inflation using mixtures of normals, while dropping the testing component for outlier detecting providing robust estimation. We introduce influence functions as a way to study the behaviour of outlying observations. Finally we discuss model fiting using JAGS.
