Keng-Hao Chang

University of Auckland, Department of Statistics

Robust regression using nonparametric scale normal mixtures

A standard assumption for linear regression is that the random noise has a normal distribution. However, for some real-world data this assumption may not necessarily hold and the model can be badly fitted, e.g., when the random noise follows a distribution with heavier tails than the normal or when there exist outliers. In this talk, I will describe a method that uses nonparametric scale normal mixtures to solve this problem.