Pearson M-Estimators in Regression Analysis.

This paper derives and adaptive partial solution for the maximum likelihood normal equations of a regression, under the assumption that the errors belong to the Pearson family. This estimator can be "robustified" producing a M-estimator with satisfactory efficiency for a wider range of error distributions. Monte-Carlo evidence on the finite sample properties of the estimates is reported. The computational requirements are very modest: all the proposed improvements can be computed with the help of an auxilliary regression.