Likelihood disparity estimation for the skew-normal distribution.

In this paper we introduce an estimator based on Likelihood Disparity which is a distance between distributions based on the Likelihood. This estimator is used to estimate location, scale and shape parameters of the skew normal distribution. Maximum Likelihood, Method of Moments and others estimation methods are often unreliable estimators of the shape parameter (they are biased towards infinity), especially for small sample size. Our estimator performs very well for all sample sizes and shape values. Its behavior is similar to the Maximum Likelihood estimator when the latter is finite, but provides a finite estimate in all remain cases