Fabrizio Leisen

Interacting multiple try algorithms with different proposal distributions

We introduce a new class of interacting Markov chain Monte Carlo (MCMC) algorithms which is designed to increase the efficiency of a modified multiple-try Metropolis (MTM) sampler. The extension with respect to the existing MCMC literature is twofold. First, the sampler proposed extends the basic MTM algorithm by allowing for different proposal distributions in the multiple-try generation step. Second, we exploit the different proposal distributions to naturally introduce an interacting MTM mechanism (IMTM) that expands the class of population Monte Carlo methods and builds connections with the rapidly expanding world of adaptive MCMC. We show the validity of the algorithm and discuss the choice of the selection weights and of the different proposals. The numerical studies show that the interaction mechanism allows the IMTM to efficiently explore the state space leading to higher efficiency than other competing algorithms.

Generalized Species Sampling Priors With Latent Beta Reinforcements

Many popular Bayesian nonparametric priors can be characterized in terms of exchangeable species sampling sequences. However, in some applications, exchangeability may not be appropriate. We introduce a novel and probabilistically coherent family of nonexchangeable species sampling sequences characterized by a tractable predictive probability function with weights driven by a sequence of independent Beta random variables. We compare their theoretical clustering properties with those of the Dirichlet process and the two parameters Poisson–Dirichlet process. The proposed construction provides a complete characterization of the joint process, differently from existing work. We then propose the use of such process as prior distribution in a hierarchical Bayes’ modeling framework, and we describe a Markov chain Monte Carlo sampler for posterior inference. We evaluate the performance of the prior and the robustness of the resulting inference in a simulation study, providing a comparison with popular Dirichlet process mixtures and hidden Markov models. Finally, we develop an application to the detection of chromosomal aberrations in breast cancer by leveraging array comparative genomic hybridization (CGH) data. Supplementary materials for this article are available online.

Conditionally identically distributed species sampling sequences

In this paper the theory of species sampling sequences is linked to the theory of conditionally identically distributed sequences in order to enlarge the set of species sampling sequences which are mathematically tractable. The conditional identity in distribution (see Berti, Pratelli and Rigo (2004)) is a new type of dependence for random variables, which generalizes the well-known notion of exchangeability. In this paper a class of random sequences, called generalized species sampling sequences, is defined and a condition to have conditional identity in distribution is given. Moreover, two types of generalized species sampling sequence that are conditionally identically distributed are introduced and studied: the generalized Poisson-Dirichlet sequence and the generalized Ottawa sequence. Some examples are discussed.

A vector of Dirichlet processes

Random probability vectors are of great interest especially in view of their application to statistical inference. Indeed, they can be used for determining the de Finetti mixing measure in the representation of the law of a partially exchangeable array of random elements taking values in a separable and complete metric space. In this paper we describe a construction of a vector of Dirichlet processes based on the normalization of completely random measures that are jointly infinitely divisible. After deducing the form of the Laplace exponent of the vector of the gamma completely random measures, we study some of their distributional properties. Our attention particularly focuses on the dependence structure and the specific partition probability function induced by the proposed vector.

Bayesian Model Selection for Beta Autoregressive Processes

We deal with Bayesian inference for Beta autoregressive processes. We restrict our attention to the class of conditionally linear processes. These processes are particularly suitable for forecasting purposes, but are difficult to estimate due to the constraints on the parameter space. We provide a full Bayesian approach to the estimation and include the parameter restrictions in the inference problem by a suitable specification of the prior distributions. Moreover in a Bayesian framework parameter estimation and model choice can be solved simultaneously. In particular we suggest a Markov-Chain Monte Carlo (MCMC) procedure based on a Metropolis-Hastings within Gibbs algorithm and solve the model selection problem following a reversible jump MCMC approach.

A multivariate extension of a vector of two-parameter Poisson-Dirichlet processes

In the big data era there is a growing need to model the main features of large and non-trivial data sets. This paper proposes a Bayesian nonparametric prior for modelling situations where data are divided into different units with different densities, allowing information pooling across the groups. Leisen and Lijoi [(2011), ‘Vectors of Poisson–Dirichlet processes’, J. Multivariate Anal., 102, 482–495] introduced a bivariate vector of random probability measures with Poisson–Dirichlet marginals where the dependence is induced through a Lévys Copula. In this paper the same approach is used for generalising such a vector to the multivariate setting. A first important contribution is the derivation of the Laplace functional transform which is non-trivial in the multivariate setting. The Laplace transform is the basis to derive the exchangeable partition probability function (EPPF) and, as a second contribution, we provide an expression of the EPPF for the multivariate setting. Finally, a novel Markov Chain Monte Carlo algorithm for evaluating the EPPF is introduced and tested. In particular, numerical illustrations of the clustering behaviour of the new prior are provided.

Asymptotic Results for a Generalized Pólya Urn with “Multi-Updating” and Applications to Clinical Trials

In this article, a new Pólya urn model is introduced and studied; in particular, a strong law of large numbers and two central limit theorems are proved. This urn generalizes a model studied in Berti et al. (2004), May et al. (2005), and in Crimaldi (2007), and it has natural applications in clinical trials. Indeed, the model includes both delayed and missing (or null) responses. Moreover, a connection with the conditional identity in distribution of Berti et al. (2004) is given.

A Bayesian Beta Markov Random Field Calibration of the Term Structure of Implied Risk Neutral Densities

We build on Fackler and King (1990) and propose a general calibration model for implied risk neutral densities. Our model allows for the joint calibration of a set of densities at different maturities and dates. The model is a Bayesian dynamic beta Markov random field which allows for possible time dependence between densities with the same maturity and for dependence across maturities at the same point in time. The assumptions on the prior distribution allow us to compound the needs of model flexibility, parameter parsimony and information pooling across densities.

A New Multinomial Model and a Zero Variance Estimation

The analysis of categorical response data through the multinomial model is very frequent in many statistical, econometric, and biometric applications. However, one of the main problems is the precise estimation of the model parameters when the number of observations is very low. We propose a new Bayesian estimation approach where the prior distribution is constructed through the transformation of the multivariate beta of Olkin and Liu (2003). Moreover, the application of the zero-variance principle allows us to estimate moments in Monte Carlo simulations with a dramatic reduction of their variances. We show the advantages of our approach through applications to some toy examples, where we get efficient parameter estimates.

An approximate likelihood perspective on ABC methods

We are living in the big data era, as current technologies and networks allow for the easy and routine collection of data sets in different disciplines. Bayesian Statistics offers a flexible modeling approach which is attractive for describing the complexity of these datasets. These models often exhibit a likelihood function which is intractable due to the large sample size, high number of parameters, or functional complexity. Approximate Bayesian Computational (ABC) methods provides likelihood-free methods for performing statistical inferences with Bayesian models defined by intractable likelihood functions. The vastity of the literature on ABC methods created a need to review and relate all ABC approaches so that scientists can more readily understand and apply them for their own work. This article provides a unifying review, general representation, and classification of all ABC methods from the view of approximate likelihood theory. This clarifies how ABC methods can be characterized, related, combined, improved, and applied for future research. Possible future research in ABC is then suggested.