BAYESIAN ANALYSIS OF FINITE MIXTURES OF WEIBULL DISTRIBUTIONS

Abstract
The paper considers Bayesian inference for finite mixtures of Weibull distributions. Despite their importance and flexibility in modeling, exact inference procedures for mixtures of Weibull distributions have not been considered before. The new methods are based on the hierarchical representation of Weibull mixtures, and computational implementation are organized around Gibbs sampling with data augmentation. The approach is illustrated using two- and three-component Weibull mixtures for the Hang-Seng stock price index.