The purpose of this paper is to present an algorithm to calculate the least median squares LMS line. In the algorithm, random strips which would catch at least fifty percent of observations are thrown on the convex hull of the data in the variable space. The LMS line lies exactly at the middle...