A REGRESSION MODEL

Abstract

In the normal linear regression the least square estimation of the coefficients has a series of nice properties. In addition the needed numerical calculations may be done a fairly efficient way. For the case when the application of this widely used least square method is not justified some further methods are suggested. But experience about their statistical and numerical properties can hardly be found. This paper intends to help the practitioner to become familiar with a type of non-least-square regression methods.