An Efficient Leave One Block Out approach to identify outliers

Abstract

In Least Squares (LS), the linearized functional model betweenM observables and N unknown parameters is given. LS provides estimates of parameters, observables, residuals and a posteriori variance. To identify outliers and to estimate accuracies and reliabilities, tests on the model and on the individual residuals can be performed at different levels of significance and power. However, LS is not robust: one outlier could be spread into all the residuals and its identification is difficult.