The Selection of an Extreme-Value Distribution and the Problem of Transforming to a Specified Distribution

摘要

We consider the problem of transforming a known distribution (or a random sample) to distributions other than the normal. Transformations to the negative exponential distribution are considered but the most interesting version is the extension concerning the selection of a best fitting extreme value distribution. By creating a suitable three-parameter family of transformations all three types of extreme value distributions may be considered as candidates. Our sample approach to the selection is based on the marginal posterior distribution of the shape parameter. (Author)