Transparent Data Mining for Big and Small Data

Abstract

The performances of three different estimators for the roughness parameterof an important distribution for synthetic aperture radar data analysisare compared: those of a moment estimator, of the maximum likelihoodestimator and of the second-order bias-corrected maximum likelihoodestimator. A Monte Carlo study is designed to perform this comparison,due to the untractability of the estimators distributions from ananalytical point of view. From this study, the use of the second-orderbias-corrected maximum likelihood estimator is suggested.