Robust PEROBLS 3 Method of Least Squares

Abstract

Outliers, or result with gross errors, are, practically, always present in regression models. Therefore, it is very important to study the regression with influence of outliers as low as possible. So, dealing with this problem I have studied a new robust Least Square (LS) Method, named after the author PEROBLS and based on the real stochastical model. Namely, the variance of an observation which has a gross error is substituted by its mean square error composed of random and gross observation error. The gross error is exactly and unbiasedly estimated in the LS model and therefore there exists only one or two adjustment iteration.