Box-Cox elliptical distributions with application (Q2312027)

The authors introduce the class of truncated elliptical distributions for multivariate data, and use this to define another new class of distributions, the elliptical Box-Cox distributions, to model multivariate positive data which is marginally skewed and may have heavy tails. The parameters in this distribution represent the scale and skewness of each of the components of the data, and association between different components. In certain special cases, these Box-Cox elliptical distributions reduce to well-known log-elliptical distributions and to symmetric Box-Cox distributions. As well as the definition of this new class of distributions, the main results of the paper are properties of these distributions, including marginal and conditional densities, and moments. The authors also discuss generation of random samples and parameter estimation using maximum likelihood, and consider an application of their model to a certain data set.