DETECTION OF LINEAR SUBSTRUCTURES IN CALIBRATION MODEL BY ROBUST APPROACH: MAXIMUM SUM OF BINARY-CODED RESIDUALS (MASBR) REGRESSION

摘要

The MASBR (maximum sum of binary-coded residuals) procedure proposed previously as a robust regression method has been modified and applied to find the linear substructures in a data set. A graphic approach is used to identify the acceptable error limit (AEL) in the MASBR algorithm. Some data sets with mixed linear substructures are interpreted by the proposed method showing satisfactory results. A MATLAB program for the MASBR algorithm is given.