Abstract:

In weighted total least squares, gross errors of observation data are not taken into consideration. In order to resolve this problem, a robust method of weighted total least squares with reweighting iteration based on IGG weighting function is proposed. Thorough experimental evaluation with a large number of simulation data and real-life data sets has been carried out. The results of coefficient estimation and error estimation are systematically compared with weighted total least squares, and robust weighted total least squares based on Huber weighting function. It is shown that (a) more reliable parameter estimation can be obtained by two robust weighted total least squares; and (b) more importantly, the proposed method performs better than others.

Key words: weighted total least squares (WTLS), reweighting iteration, robust WTLS, Huber weighting function, IGG weighting function