Wavelet Estimator in Nonparametric Regression to improve Least Squares Estimation

A nonparametric regression with a wavelet estimator is proposed to correct undesired effects
(bias) that, in general, can not be handle in functional or stochastical models of Least Squares method. An application was carried out with real data from Global Positioning System to correct the multipath effect from signal reflection. Improviments of up to 99% was reached in root mean squared error of the residuals.