Skew-t Information Matrix: Evaluation and Use

Abstract

Azzalini’s skew-t distributions, described in detail in Azzalini [2013], have become very popular because of their practical usefulness and the complete R package sn for skew-normal distributions that include extensive support for fitting and analysis of skew-t distributions. A major difficulty is that the skew-t distribution expected information matrix has no analytical form. This thesis develops an R package skewtInfo to compute the expected information matrix through numerical integrations and investigates the accuracy of the resulting matrix and its use for computing finite-sample approximations for skew-t parameter estimates.