Abstract

The least median of squares (LMS) regression method (Rousseeuw 1984) is highly robust to outliers in the data. It can be computed by means of PROGRESS (from Program for RObust reGRESSion) described in (Rousseeuw and Leroy 1987). After ten years we have developed a new version of PROGRESS, which also computes the least trimmed squares (LTS) method. We will discuss the various new features of PROGRESS, with emphasis on the algorithmic aspects.

Keywords

Algorithm, Breakdown value, Least Median of Squares, Least Trimmed Squares, Robust regression.