Advances in finite mixture models:new methodologies and applications

The extent of finite mixture models has widened considerably over the last century, from both a theoretical and a practical point of view. Their usefulness and flexibility is discussed in this thesis, which consists of a collection of four manuscripts that have as common background new methodologies and applications of finite mixture models. The first two manuscripts focus on specific economics and financial topics, and the finite mixture models are mainly used as a mathematical device for obtaining a flexible and tractable density. This has important consequences for the estimation of some commonly used risk measures. The other two manuscripts aim to use finite mixture models for clustering in a matrix-variate framework. In all the manuscripts, parameter estimation is carried by using the maximum-likelihood approach, implemented directly or via variants of the expectation-maximization algorithm. Both simulated and real datasets are used for illustrative purposes in each manuscript.