Parameter Estimation in a Three-Parameter Lognormal Distribution

Abstract

Point estimation methods for the three-parameter lognormal distribution are investigated and compared. The lognormal distribution is required in many topical areas, but so far there have been no comparative studies of the various estimation methods. We show that despite the large number of traditional estimation methods, the lognormal distribution requires special methods. We accordingly consider specializations of the main parameter estimation approaches, including the actively developing distance minimization methods. Their accuracy and speed are compared on simulated data. We show that specialized parameterestimation methods may outperform the highly popular maximum likelihood method.