Sir David F. Hendry Recommends “An Analysis of the Indicator Saturation Estimator as a Robust Regression Estimator” by Søren Johansen and Bent Nielsen

Abstract

It may be thought that testing for outliers and shifts everywhere in a sample might adversely affect statistical inference. Fortunately, the rigorous and innovative analysis by Søren Johansen and Bent Nielsen for impulse-indicator saturation (IIS) allays such concerns. Under the null of no outliers, the limiting distribution of the IIS estimator of economic parameters of interest converges to their population parameters at the usual rate and is normal for stationary data. Yet IIS checks for an unknown number of outliers, of unknown magnitudes and signs, not knowing in advance where they occur in the data.