Computational Comparisons of Two-Component Mixtures Using Lindley-Type Models

The Lindley distribution is well known as a suitable, parsimonious model for analyzing lifetime data. Its simplicity however, can limit its applicability in certain cases, thus motivating the need to enhance its flexibility. Through the years, many authors have put a twist on the Lindley distribution, where one of these extensions includes a mixture modeling construction. In this chapter, we provide a computational exploration of finite mixtures of Lindley-type models that include, as components, well-known cases from literature as well as a proposed alternative parsimonious mixture model. Simulation studies investigate the effect and degree of the additional parameter(s) of the finite mixtures of Lindley-type models, when compared to the finite mixture of the one parameter Lindley model. The EM algorithm is considered in this study as a suitable method for the parameter estimation. The practical implementation of these different mixture models is illustrated through real data examples, with observations coming from a survival analysis domain.