APPLICATION OF L-CURVE METHOD TO EQUIVALENTWEIGHT ROBUST RIDGE ESTIMAION MODEL
Wang Bin 1, 2) ; Gao Jingxiang 1) ; Liu Chao 1, 2) ; Wang Jian 1) ; and Zhou Feng 1, 2)
1)China University of Mining and Technology, CESSI, Xuzhou 221116;2)Jiangsu Key Laboratory of Resources and Environmental Information Engineering, Xuzhou 221116
Abstract An equivalent weight robust ridge estimation model is derived according to the idea of robust estimation for correlated observations, and then the Lcurve method is introduced to this model for the solution of ridge parameters. In the iterative process, equivalent weight matrix may be non positive definite so it can not be unitized through classical Cholesky decomposition. According to the characteristic of bifactor model, an algorithm of equivalent weight matrix decomposition is proposed to solve this problem, then the Lcurve method can be successfully used in this process. The calculated example manifests that this model can effectively eliminate the influence of coefficient matrix’s morbidity and gross errors.