Small sample properties of the maximum likelihood estimators for an alternative parameterization of the three−parameter lognormal distribution

摘要: The small sample properties of the score function approximation to maximum likelihood estimator for three-parameter lognormal distribution using an alternative parameterization are considered. new set parameters is a continuous usual parameters. However, unlike with parameterization, technique this extremely insensitive starting values. Further, it shown that whenever third moment less than zero, local exists at boundary point. For point unattainable. parameter space can be expanded include these points. This procedure results in good estimates expected value, variance, extreme percentiles and other even samples where, typical estimation fails converge.