The Least Absolute Deviation Estimation Method of Parameter in Nonlinear Regression Model and Its Feasible Direction Algorithm

摘要

For parameter estimation methods in regression models, the least absolute deviation estimation is more robust than the least square sum estimation in many cases. However, the computation of the least absolute deviation estimation is difficult. In this paper, we reduce the nonlinear least absolute deviation estimation problem to nonlinear programming problem and design a feasible direction algorithm to solve the nonlinear least absolute deviation estimation problem. Numerical example demonstrating the efficiency of this algorithm is given.