A1260
Title: Confidence intervals based on L-moments for population quantiles of the three-parameter kappa distribution
Authors: Tereza Slamova - Technical University of Liberec (Czech Republic) [presenting]
Abstract: Conventional moments are traditionally used to describe features of a univariate probability distribution. An alternative approach uses quantities called L-moments, which are based on a linear combination of the expected value of the order statistics. Their main advantages in comparison to the conventional moments are the existence of all orders under only a finite mean assumption and higher robustness to outliers. L-moments also provide parameter estimators in a similar way as in the moment's method, and moreover, the method based on them is in some cases preferred over traditional methods for quantile estimation, specifically when such distributions with heavier tails than the normal distribution are fitted to smaller samples. A three-parameter kappa distribution is a special case of the four-parameter kappa distribution, and it has a heavier tail. The asymptotic confidence intervals for population parameters and quantiles of the three-parameter kappa distribution have been derived by using L-moments and compared to those obtained by conventional moments and maximum likelihood methods via Monte Carlo simulations.
