Variation characteristics of MDBs in robust estimation

Autor(en): Jianteng Guo, Jikun Ou

The least-squares (LS) method is very susceptive to outliers. The effects of any perturbation of tue variance components and weight elements on tue LS method are analyzed and tested. A local sensitivity-based robust estimation is developed. Theoretical analyses and numerical results indicate that the MDB (minimal detectable bias) measures become larger and larger with Progress in the iterative process, thus confirming the robustness of the proposed method.