Algorithm for estimates determination of parameters of regression straight polyline with objective elimination of outliers

This paper deals with the algorithm development for estimate calculation of parameters of regressive straight polyline of the measured y = f(x) dependence, for example linear parts of dependence of temperature as a function of heat during melting. The regression straight polyline is determined by segments of the straight lines. The algorithm input data for regression straight lines calculation are experimentally measured sets of points in [x,y] coordinates. The regression straight lines equations are given by using the least-square method. The regression line equations are calculated with respect to the elimination of outliers. A decision, whether or not the measured point of functional dependence is remote from the linear segment, is done in an objective way on the basis of a statistical criterion, which is a part of the algorithm. The algorithm finishes by calculating parameters of the corresponding regression straight polyline and other statistical characteristics. On the basis of the suggested algorithm, a program has been assembled, the reliability of which was verified on a serious model and through practical use.