Inferring components and clusters in Bayesian finite mixture modelling

Beschreibung

An issue of finite mixture modelling is the specification of an appropriate number of mixture components. The choice of the number of mixture components is crucial, since it is assumed that each mixture component should capture exactly one data cluster. In the Bayesian framework, a natural approach to estimate the number of components is to treat it as an unknown model parameter and to specify a prior on it. Several inference methods have been proposed for estimating this model, in particular, reversible jump Markov chain Monte Carlo (RJMCMC) techniques. However, it can be a difficult task to design suitable proposal densities in higher-dimensional parameter spaces. In our approach we distinguish between the total number of mixture components in the probability distribution of the mixture model and the number of 'active' components, i.e. components, to which observations are assigned and which correspond to the data clusters. Using recently published results by Miller and Harrison (2016), we are able to make inference on both the number of mixture components and the number of data clusters without employing complex transdimensional sampling techniques, rather using the familiar Gibbs sampling. The method is illustrated using simulation studies and real applications.

Zeitraum	5 Juli 2017
Ereignistitel	Statistics4@Florence
Veranstaltungstyp	Keine Angaben
Bekanntheitsgrad	International