Extreme value analysis of diamond-size distributions

Extreme value analysis provides a semiparametric method for analyzing the extreme long rails of skew distributions which may be observed when handling mining darn. The estimation of important tail characteristics, such as the extreme value index, allows for a discrimination between competing distribution models. It measures the ''thickness'' of such long railed distributions, if only a limited sample is available. This paper stresses the practical implementation of extreme value theory, which is used to discriminate a lognormal from a mixed lognormal distribution in a case study of size distributions for alluvial diamonds.