Sensitivity to gross errors of new variants of MSPLIT estimators

Abstract

This study examines the sensitivity of shift estimators between competitive parameters in the Msplit estimation method to potential disturbances in observational data caused by gross errors. The essence of parameter estimation in a split functional model of geodetic observations is to determine, based on a single set of measurements, competitive values of the unknown parameter vector. In this way, the approach naturally serves as a method for determining parameter vector displacements, for example in the deformation analysis of geodetic networks. Using a Monte Carlo approach, empirical influence functions were obtained for the classical squared Msplit estimators of shift of parameters as well as for new variants of this method, including those based on the concept of a strengthening matrix for measurement results corresponding to a specific competing functional model and the elimination of the reversal point effect. The empirical analyses were conducted using an example concerning the determination of displacements of controlled points in a leveling control network. The obtained results were compared with those produced using the classical least-squares method.