# Seminars in Statistics

## March 2019

### Jouchi Nakajima (Bank for International Settlements (BIS))

"Effectiveness of unconventional monetary policies in a low interest rate environment"

Find out more »## December 2018

### Matteo Sesia (Stanford University)

"New tools for reproducible variable selection with knockoffs"

Find out more »### Gary L. Rosner (Johns Hopkins University)

"Bayesian Approaches in Regulatory Science"

Find out more »## November 2018

### Eleni Matechou (University of Kent)

"Bayesian nonparametric modelling of phenology using capture-recapture data"

Find out more »## October 2018

### Eduard Belitser (Vrije Universiteit Amsterdam)

Robust inference for general projection structures by empirical Bayes and penalization methods

Find out more »### Stephanie van der Pas (Leiden University)

Posterior concentration for Bayesian regression trees and their ensembles

Find out more »### Fernando A. Quintana (Pontificia Universidad Catolica de Chile)

Discovering Interactions Using Covariate Informed Random Partition Models

Find out more »## August 2018

### Jan Naudts (Universiteit Antwerpen)

Non-Commutative Information Geometry Information geometry is concerned with the study of statistical manifolds. These are differentiable manifolds consisting of probability distributions. In the param- eterized case their geometry is described by a metric tensor and a pair of dually flat connections. In the more general non-parameterized case they are Banach manifolds. This area of research is still developing and has applications in many domains. My interest in this domain is twofold. The notion of an exponential family of statistical…

Find out more »## May 2018

### Kolyan Ray (King’s College London)

Estimating the mean response in a missing data model We study semiparametric Bayesian estimation of the mean response in a binary regression model with missing observations. We allow some dependence between the missingness and response mechanisms, which we assume are conditionally independent given some measured covariates (i.e. unconfoundedness). This model has applications in biostatistics and causal inference. We show that the marginal posterior distribution for the mean response arising from product priors on the different model parameters satisfies a semiparametric Bernstein-von…

Find out more »## April 2018

### Theodore Kypraios (University of Nottingham)

Latent Branching Trees: Modelling and Bayesian Computation. In this talk a novel class of semi-parametric time series models will bepresented, for which we can specify in advance the marginal distributionof the observations and then build the dependence structure of theobservations around them by introducing an underlying stochastic processtermed as 'latent branching tree'. It will be demonstrated how can wedraw Bayesian inference for the model parameters using Markov ChainMonte Carlo methods as well as Approximate Bayesian Computationmethodology. Finally a real dataset…

Find out more »### Brunero Liseo (Università di Roma La Sapienza)

Modelling Preference Data with the Wallenius Distribution The Wallenius distribution is a generalisation of the Hypergeometric distribution where weights are assigned to balls of different colours. This naturally defines a model for ranking categories which can be used for classification purposes. Since, in general, the resulting likelihood is not analytically available, we adopt an approximate Bayesian computational (ABC) approach for estimating the importance of the categories. We illustrate the performance of the estimation procedure on simulated datasets. Finally, we use…

Find out more »## February 2018

### Fabrizio Leisen (University of Kent)

Compound Random Measures Compound Random Measures (CoRM's) have been recently introduced by Griffin and Leisen (2017) and represent a general and tractable class of vectors of Completely Random Measures. This talk aims to provide an overview about CoRM's by illustrating some recent developments about their use in Bayesian nonparametrics.

Find out more »## January 2018

### Davide La Vecchia (University of Geneva)

Saddlepoint techniques for dependent data Saddlepoint techniques provide numerically accurate, higher-order, small sample approximations to the distribution of estimators and test statistics. While a rich theory is available for saddlepoint techniques in the case of independently and identically distributed observations, only a few results have been obtained for dependent data. In this talk, we explain how to fill this gap in the literature. Using the method of the tilted-Edgeworth expansion, we devise new saddlepoint density approximations and saddlepoint test statistics…

Find out more »## December 2017

### John Armstrong (King’s College London)

Stochastic Differential Equations as Jets We explain how Ito Stochastic Differential Equations (SDEs) on manifolds may be defined using 2-jets of smooth functions. We show how this relationship can be interpreted in terms of a convergent numerical scheme. We show how jets can be used to derive graphical representations of Ito SDEs and how jets give rise to a coordinate free approach to understanding SDEs and diffusions on manifolds. We will consider some applications of this approach.

Find out more »## November 2017

### Ester Mariucci (Humboldt-Universität zu Berlin)

Wasserstein distances and other metrics for discretely observed Lévy processes We present some upper bounds for the Wasserstein distance of order p between the product measures associated with the increments of Lévy processes with possibly infinite Lévy measures. As an application, we derive an upper bound for the total variation distance between the marginals of Lévy processes with possibly infinite Lévy measures and non-zero Gaussian components. A lower bound for the Wasserstein distance of order p is also presented. Furthermore,…

Find out more »### Krzysztof Łatuszyński (University of Warwick)

Exact Bayesian inference for discretely observed jump-diffusions The standard approach to inference for parametric diffusion processesrelies on discretisation techniques (such as the Euler method) thatintroduce an approximation error difficult to quantify especially fordiscontinuous models, like jump-diffusions.In this talk, I will present methodology for exact inference thatavoids discretisation errors and allows to design MCMC samplerstargeting the exact posterior distribution of the diffusion parametersand diffusion path between observations. The approach is based onBernoulli Factory type subroutines, and is a general alternative topseudo-marginal…

Find out more »## October 2017

### Kazuhiko Kakamu (Kobe University)

How does monetary policy affect income inequality in Japan? Evidence from grouped data Co-author: Martin Feldkircher (Oesterreichische Nationalbank (OeNB)) Abstract: We examine the effects of monetary policy on income inequality in Japan using a novel econometric approach that jointly estimates the Gini coefficient based on micro-level grouped data of households and the dynamics of macroeconomic quantities. Our results indicate different effects on income inequality depending on the monetary policy measure under consideration: A traditional rate increase decreases income inequality, whereas…

Find out more »### Julien Berestycki (University of Oxford)

Branching Brownian motion with absorption What does the genealogy of a population under selection look like? This question is crucial for ecology and evolutionary biology and yet it is not fully understood. Recently, Brunet and Derrida have conjectured that for a whole class of models of such populations, we can expect the genealogy to be described by a universal scaling limit: the Bollthausen-Sznitman coalescent. The purpose of this talk is to present several recent results which put this prediction on a rigorous…

Find out more »## May 2017

### Petros Dellaportas (University College London)

High dimensional jump processes with stochastic volatility We deal with the problem of identifying jumps in multiple financial time series using the stochastic volatility model combined with a jump process. We develop efficient MCMC algorithms to perform Bayesian inference for the parameters and the latent states of the proposed models. In the univariate case we use an homogeneous compound Poisson process for the modelling of the jump component. In the multivariate case we adopt an inhomogeneous Poisson process, with intensity…

Find out more »## April 2017

### John Aston (University of Cambridge)

Functions, Manifolds and Statistical Linguistics Functional Data Analysis concerns the statistical study of curves and surfaces. An extension, Functional Object Data Analysis, looks at the statistical analysis of curves and surfaces which live in restricted spaces, such as on manifolds. One particular example of these are the covariance functions associated with the underlying curves. These positive definite operators can often be of interest in their own right, yet standard functional data models are not appropriate due to the positive definite…

Find out more »## March 2017

### Maria De Iorio (University College London)

Dependent Generalised Dirichlet Process Priors We propose a novel Bayesian nonparametric process prior for modelling a collections of random discrete distributions. This process is defined by combining a Generalised Dirichlet Process with a suitable Beta regression framework that introduces dependence among the discrete random distributions. This strategy allows for covariate dependent clustering of the observations. Some advantages of the proposed approach include wide applicability, ease of interpretation and efficient MCMC algorithms. The methodology is illustrated through two real data applications…

Find out more »## January 2017

### Stéphane Boucheron (Université Paris-Diderot)

Concentration inequalities in the infinite urn scheme for occupancy counts and the missing mass, with applications to Good-Turing estimators and adaptive statistical text compression An infinite urn scheme is defined by a probability mass function over positive integers. A random allocation consists of a sample of N independent drawings according to this probability distribution where N may be deterministic or Poisson-distributed. We are concerned with occupancy counts, that is with the number of symbols with r or at least r…

Find out more »## December 2016

### Vinayak Rao (Purdue University)

Path and parameter inference for Markov jump processes A variety of phenomena are best described using dynamical models which operate on a discrete state-space and in continuous time. The most common example is the Markov jump processes whose applications range from systems biology, genetics, computing networks and human-computer interactions. Posterior computations typically involve approximations like time discretization and can be computationally intensive. In the first half of this talk, I will describe some previous work (joint with Yee Whye Teh)…

Find out more »### Tamara Broderick (MIT)

Fast Quantification of Uncertainty and Robustness with Variational Bayes In Bayesian analysis, the posterior follows from the data and a choice of a prior and a likelihood. These choices may be somewhat subjective and reasonably vary over some range. Thus, we wish to measure the sensitivity of posterior estimates to variation in these choices. While the field of robust Bayes has been formed to address this problem, its tools are not commonly used in practice---at least in part due to…

Find out more »### Alejandro Jara (Pontificia Universidad Católica de Chile)

Bayesian nonparametric approaches for the analysis of compositional data based on Bernstein polynomials We will discuss Bayesian nonparametric procedures for density estimation and fully nonparametric regression for compositional data, that is, data supported in a multidimensional simplex. The procedures are based on modified classes of Bernstein polynomials. We show that the modified classes retain the well known approximation properties of the classical versions defined on an hypercube and on a multidimensional simplex. Based on these classes, we define prior distributions,…

Find out more »## November 2016

### Bruno Scarpa (University of Padua)

Bayesian modelling of networks in business intelligence problems Complex network data problems are increasingly common in many fields of application. Our motivation is drawn from strategic marketing studies monitoring customer choices of specific products, along with co-subscription networks encoding multiple purchasing behavior. Data are available for several agencies within the same insurance company, and our goal is to efficently exploit co-subscription networks to inform targeted advertising of cross-sell strategies to currently mono-product customers. We address this goal by developing a…

Find out more »## September 2016

### Boyu Ren (Harvard T.H. Chan School of Public Health)

A Bayesian Nonparametric model for microbiome data analysis We develop a statistical model to analyse microbiome profiling data based on sequencing of genetic fingerprints in 16S ribosomal RNA. The analysis allows us to quantify the uncertainty in ecological ordination and clustering methods commonly applied in microbiome research. In addition, it can be extended into a framework for association studies when sample characteristics are available. The method is based on the estimation of the underlying microbial distribution in experimental samples using…

Find out more »## June 2016

### Steven Scott (Google)

Predicting the Present with Bayesian Structural Time Series This article describes a system for short term forecasting based on an ensemble prediction that averages over different combinations of predictors. The system combines a structural time series model for the target series with regression component capturing the contributions of contemporaneous search query data. A spike-and-slab prior on the regression coefficients induces sparsity, dramatically reducing the size of the regression problem. Our system averages over potential contributions from a very large set…

Find out more »### Natesh Pillai (Harvard University)

Bayesian Factor Models in High Dimensions Sparse Bayesian factor models are routinely implemented for parsimonious dependence modeling and dimensionality reduction in high-dimensional applications. We provide theoretical understanding of such Bayesian procedures in terms of posterior convergence rates in inferring high-dimensional covariance matrices where the dimension can be larger than the sample size. We will also discuss other high dimensional shrinkage priors and discuss them in the context of factor models.

Find out more »## May 2016

### Richard Nickl (University of Cambridge)

Nonparametric Bayesian inference for discretely sampled diffusions We consider the nonlinear statistical inverse problem ofmaking inference on the unknown parameters of a diffusion processdescribing the solution of a stochastic differential equation. Theobservation regime is such that the process is sampled at discretetime points that are a fixed distance apart, and we investigate theasymptotic regime when more samples accrue in the time horizon (thusavoiding unrealistic `high frequency’ assumptions). We shall brieflyreview frequentist estimation techniques and then turn to Bayesiannonparametric approaches to…

Find out more »## April 2016

### Laura Ventura (University of Padua)

Robust Approximate Bayesian Inference The likelihood function is the basis of both frequentist and Bayesian methods. However, the stability of likelihood-based procedures requires strict adherence to the model assumptions: mild deviations from the model can lead to misleading inferential results. A possible Bayesian solution to robustness is to use a of robust pseudo likelihood, such as the quasi- and the empirical likelihoods, in the Bayes formula. However, in multiparameter problems the quasi-likelihood is cumbersome, while with small samples the empirical…

Find out more »## March 2016

### Emilie Kaufmann (CNRS, France)

The information complexity of sequential resource allocation I will talk about sequential resource allocation, under the so-called stochastic multi-armed bandit model. In this model, an agent interacts with a set of (unknown) probability distributions, called 'arms' (in reference to 'one-armed bandits', another name for slot machines in a casino). When the agent draws an arm, he observes a sample from the associated distribution. This sample can be seen as a reward, and the agent then aims at maximizing the sum…

Find out more »## December 2015

### Mattia Ciollaro (Carnegie Mellon University)

An inferential theory of clustering for functional data Recently, it has been shown that Morse theory can be exploited to de- velop a sound inferential background for clustering: one can rigorously define both population and empirical clusters by means of the gradient flows asso- ciated to the population density p and the estimated density pˆ. In this framework, clusters are well-defined entities corresponding to the basins of attraction of the density’s critical points. The population parameter of inter- est corresponds…

Find out more »### Juhee Lee (University of California at Santa Cruz)

Bayesian inference for intra-tumor heterogeneity in mutations and copy number variation Tissue samples from the same tumor are heterogeneous. They consist of different subclones that can be characterized by differences in DNA nucleotide sequences and copy numbers on multiple loci. Inference on tumor heterogeneity thus involves the identification of the subclonal copy number and single nucleotide mutations at a selected set of loci. We carry out such inference on the basis of a Bayesian feature allocation model. We jointly model…

Find out more »### Harry Crane (Rutgers University)

Relative exchangeability Symmetry arguments lie at the heart of classical considerations in inductive inference and statistics. In statistics, de Finetti's notion of exchangeability is the most prominent symmetry assumption, laying the foundation for Bayesian inference. In practice, many statistical and scientific problems exhibit only partial symmetry determined by some underlying structure in a population. As a simple example, consider measurements X_1, X_2, ... taken on a population of men and Y_1, Y_2, ... taken on a population of women. Without…

Find out more »## November 2015

### Paul Jenkins (University of Warwick)

Exact simulation of the Wright-Fisher diffusion The Wright-Fisher family of diffusion processes is a class of evolutionary models widely used in population genetics, with applications also in finance and Bayesian statistics. Simulation and inference from these diffusions is therefore of widespread interest. However, simulating a Wright-Fisher diffusion is difficult because there is no known closed-form formula for its transition function. In this talk I show how it is possible to simulate exactly from the scalar Wright-Fisher diffusion with general drift,…

Find out more »## October 2015

### David Rossell (University of Warwick)

Consistency of posterior model probabilities in high-dimensional model selection In recent years there has been an increasing interest in developing Bayesian formulations that remain effective in high-dimensional and non-standard problems. We focus on high-dimensional model selection problems where the number of parameters may grow with the sample size, and review results characterizing the situations under which one can hope for the posterior probability assigned to the data-generating truth to eventually converge to 1 (often termed "strong consistency"). As it turns…

Find out more »### Frédéric Lavancier (Université de Nantes)

Determinantal point process models and statistical inference In this talk, I will demonstrate that Determinantal point processes (DPPs) provide useful models for the description of repulsive spatial point processes. Such data are usually modeled by Gibbs point processes, where the likelihood and moment expressions are intractable and simulations are time consuming. I will recall the definition of a DPP and review some of these appealing properties which make DPP models well suited for statistical analysis. In particular the ’most repulsive’…

Find out more »## September 2015

### Mingyuan Zhou (University of Texas at Austin)

The Poisson gamma belief network A key issue in deep learning is to define an appropriate network structure, including both the depth of the network and the width of each hidden layer, which may be naturally addressed with completely random measures. We propose the Poisson gamma belief network (PGBN), which factorizes each of its layers into the product of a connection weight matrix and the nonnegative real hidden units of the next layer, to infer a multilayer representation of high-dimensional…

Find out more »## May 2015

### Luca Tardella (University of Rome “La Sapienza”)

Flexible behavioral capture-recapture modelling We develop some new strategies for building and fitting new flexible classes of para- metric capture-recapture models for closed populations which can be used to address a better understanding of behavioural patterns. We first rely on previous approaches based on a conditional probability parameterization and review how to regard a large subset of standard capture-recapture models as a suitable partitioning in equivalence classes of the full set of conditional probability parameters. We then propose a regression…

Find out more »### Andreas Kyprianou (University of Bath)

Deep factorisation of stable processes The Lamperti-Kiu transformation for real-valued self-similar Markov processes (rssMp) states that, associated to each rssMp via a space-time transformation, is a Markov additive process (MAP). In the case that the rssMp is taken to be an α-stable process with α∈(0,2), the characteristics of the matrix exponent of the semi-group of the embedded MAP (the Lamperti-stable MAP) is known. Specifically, the matrix exponent of the Lamperti-stable MAP’s transition semi-group can be written in a compact form…

Find out more »## March 2015

### Piotr Zwiernik (University of Genova)

Maximum likelihood estimation for linear Gaussian covariance Models We study parameter estimation in linear Gaussian covariance models, which are p-dimensional Gaussian models with linear constraints on the covariance matrix. Maximum likelihood estimation for this class of models leads to a non-convex optimization problem which typically has many local optima. We prove that the log-likelihood function is concave over a large region of the cone of positive definite matrices. Using recent results on the asymptotic distribution of extreme eigenvalues of the…

Find out more »### Luigi Malagò (Shinshu University, Japan)

Information geometry of the Gaussian distribution in view of stochastic optimization: first and second order geometry We study the optimization of a continuous function by its stochastic relaxation, i.e., the optimization of the expected value of the function itself with respect to a density in a statistical model. In the first part of the talk we focus on gradient descent techniques applied to models from the exponential family and in particular on the multivariate Gaussian distribution. From the theory of…

Find out more »## February 2015

### Alessandro Arlotto (Duke University)

Sequential decisions, time dependence, and central limit theorems We prove a central limit theorem for the sum of functions of (1+m)-dimensional vectors from a time non-homogeneous Markov chain and we show several examples in which this central limit theorem can be used to easily establish the asymptotic normality of the optimal total reward of finite horizon Markov decision problems. By choosing m=0 we recover the classic central limit theorem of Dobrushin (1956). Joint work with J.M. Steele.

Find out more »## January 2015

### Jean-Bernard Salomond (CWI, Netherlands)

Bayesian nonparametric testing for embedded hypotheses with application to shape constrains If Bayesian nonparametric methods have received a great interest in the literature, only a few is known for testing nonparametric hypotheses, and especially the asymptotic properties of such tests. The problem of testing between two nonparametric hypotheses is known to be difficult, but the problem becomes even harder when the hypotheses are embedded. In this work, we propose a method to circumvent these difficulties with a special focus on…

Find out more »## December 2014

### Luis Enrique Nieto-Barajas (ITAM, México)

Spatial gamma processes in disease mapping In this talk we will present Bayesian models based on Markov random fields of gamma type to model the relative risk in disease mapping data. The spatial gamma processes allow for different spatial dependence among neighbours. We describe the properties of ths processes and use them as prior distributions to carry out Bayesian inference. We extend the model in two ways: to include covariates, and to allow for excess of zeroes in the data.…

Find out more »### Jaeyong Lee (Seoul National University)

Dependent species sampling models We consider a novel Bayesian nonparametric model for density estimation with an underlying spatial structure. The model is built on a class of species sampling models, which are discrete random probability measures that can be represented as a mixture of random support points and random weights. Specifically, we construct a collection of spatially dependent species sampling models and propose a mixture model based on this collection. The key idea is the introduction of spatial dependence by…

Find out more »### Taeryon Choi (Korea University)

Generalized partially additive Bayesian spectral analysis regression models In this talk, we present a Bayesian method for generalized partially additive regression using a spectral analysis of Gaussian process priors for the regression function. The smoothing prior distribution for the spectral coefficients incorporates hyper parameters that control the smoothness of the function and the tradeoff between the data and the prior distribution. We contrast our approach with existing Bayesian regression models for dealing with shape restrictions for the regression function and…

Find out more »### James Scott (University of Texas at Austin)

False discovery rate smoothing Many approaches for multiple testing begin with the assumption that all tests in a given study should be combined into a global false-discovery-rate analysis. But this may be inappropriate for many of today's large-scale screening problems, where test statistics have a natural spatial lattice structure (voxels in the brain, distance along the chromosome), and where a combined analysis can lead to poorly calibrated error rates. To address this problem, we introduce an approach called false-discovery-rate smoothing. …

Find out more »## November 2014

### Alexandros Beskos (University College London)

SMC Samplers for Applications in High Dimensions Sequential Monte Carlo (SMC) methods are nowadays routinely applied in a variety of complex applications: hidden Markov models, dynamical systems, target tracking, control problems, just to name a few. Whereas SMC methods have been greatly refined in the last decades and are now much better understood, they are still known to suffer from the curse of dimensionality: algorithms can sometimes break down exponentially fast with the dimension of the state space. As a…

Find out more »## October 2014

### Bartek Knapik (VU Amsterdam)

Convergence rates of posterior distributions in nonparametric inverse problems Since the seminal works of Ghosal, Ghosh and van der Vaart (2000) and Shen and Wasserman (2001), posterior contraction has attracted much attention, resulting in the rich literature on this subject. However, these results are not suitable to deal with trully ill-posed inverse problems, where one is interested in the parameter of interest f, given noisy observations of its transformed version Kf. General theorems yield contraction results in some metric measuring…

Find out more »## July 2014

### Michael J. Daniels (University of Texas at Austin)

A Flexible Bayesian Approach to Monotone Missing Data in Longitudinal Studies with Informative Missingness with Application to An Acute Schizophrenia Clinical Trial We develop a Bayesian nonparametric model for a longitudinal response in the presence of nonignorable missing data. Our general approach is to first specify a {em working model} that flexibly models the missingness and full outcome processes jointly. We specify a Dirichlet process mixture of missing at random (MAR) models as a prior on the joint distribution of…

Find out more »## June 2014

### Goran Peskir (University of Manchester)

Optimal Mean-Variance Portfolio Selection I will present a dynamic formulation of the mean-variance portfolio selection problem and discuss possible ways of solving it.Joint work with J. L. Pedersen (Copenhagen)

Find out more »## April 2014

### Judith Rousseau (Université Paris Dauphine)

Behaviour of the posterior distribution in HMM models when the number of states is misspecified In this paper we study the asymptotic behaviour of the posterior distribution for parametric HMM models with finite number of components. We concentrate in particular on the case where the number of states of the hidden Markov chain in the model is larger than the true number of states. In such cases the parameter is not identifiable. We study the impact of the prior on…

Find out more »## March 2014

### Dario Spanò (University of Warwick)

On the ancestral process of long-range seed bank models It has been observed that, in some bacterial species, spores may remain dormant for a long time, to wake up much later, even up to "order of population size" generations later. When they wake up, they can still participate in the population's reproduction. This incredibly relaxed attitude causes a relaxation of the population's Markov property, forward in time. I will describe some results about the genealogical process of seed bank models…

Find out more »### Christina Goldschmidt (University of Oxford)

The scaling limit of the minimum spanning tree of the complete graph Consider the complete graph on n vertices with independent and identically distributed edge-weights having some absolutely continuous distribution. The minimum spanning tree (MST) is simply the spanning subtree of smallest weight. It is straightforward to construct the MST using one of several natural algorithms. Kruskal's algorithm builds the tree edge by edge starting from the globally lowest-weight edge and then adding other edges one by one in increasing…

Find out more »## February 2014

### Nicolas Chopin (ENSAE, France)

Sequential Quasi Monte Carlo We develop a new class of algorithms, SQMC (Sequential Quasi Monte Carlo), as a variant of SMC (Sequential Monte Carlo) based on low-discrepancy points. The complexity of SQMC is O(N log N) where N is the number of simulations at each iteration, and its error rate is smaller than the Monte Carlo rate O(N^{-1/2}). The only requirement to implement SQMC is the ability to write the simulation of particle x_t^n given x_{t-1}^n as a deterministic function…

Find out more »## January 2014

### Hugo Maruri-Aguilar (Queen Mary University of London)

The algebraic method in experimental designs This seminar is part of the 3rd Carlo Alberto Stochastics Workshop

Find out more »### Jim Q. Smith (University of Warwick)

Chain event graphs and the geometry of causation This seminar is part of the 3rd Carlo Alberto Stochastics Workshop

Find out more »## December 2013

### Bas Kleijn (University of Amsterdam)

Testability and consistency Bayesian consistency theorems come in (at least) three distinct types, e.g. Doob's prior-almost-sure consistency on Polish spaces, Schwartz's Hellinger consistency with KL-priors and the `tailfree' weak consistency of Dirichlet posteriors. In this talk we ask the question how these notions of convergence are related and argue that one characterises them most conveniently using test sequences. We investigate the differences between uniform, point-wise and almost-sure testing, prove Doob-type posterior consistency theorems in various model topologies and consider hypothesis…

Find out more »### Li Ma (Duke University)

Adaptive testing of conditional association through recursive mixture modeling In many case-control studies, a central goal is to test for association or dependence between the predictors and the response. Relevant covariates must be conditioned on to avoid false positives and loss in power. Conditioning on covariates is easy in parametric frameworks such as the logistic regression—by incorporating the covariates into the model as additional variables. In contrast, nonparametric methods such as the Cochran-Mantel-Haenszel test accomplish conditioning by dividing the data…

Find out more »### Yongdai Kim (Seoul National University)

Deviance Information Criteria for the frailty model We are concerned with model selection for the frailty model by use of the deviance information criterion (DIC). The DIC is a Bayesian model selection criterion proposed by Spiegelhalter et al. (2002). A difficulty in applying the DIC to the frailty model lies on the unspecified baseline hazard function. While the DIC has been studied mostly for parametric models, it is unclear how the DIC is defined for semiparametric models. We propose various…

Find out more »## November 2013

### François Caron (University of Oxford)

A Bayesian nonparametric model for undirected and multi-edges networks In this talk, I will present ongoing work on a Bayesian nonparametric specification for either undirected or multi-edge directed networks, building on the framework of completely random measures. The formulation allows for an unbounded number of nodes in the network, while encouraging a sparse set of interactions. Importantly, as demonstrated empirically, the model captures salient features of real world networks like power-law degree distributions with possible exponential cut-off in the tails.…

Find out more »## October 2013

### Fancisco Javier Rubio (University of Warwick)

Bayesian inference in two–piece and skew–symmetric distributions using Jeffreys priors We study the Jeffreys prior and the independence Jeffreys prior of general classes of univariate location–scale two–piece and skew–symmetric models. For the case of two– piece models, Jeffreys priors are shown not to allow for Bayesian inference in the wide and practically relevant class of distributions obtained by skewing scale mixtures of normals. Easily checked conditions under which independence Jeffreys priors can be used for valid inference are derived. In…

Find out more »## July 2013

### Andrés Felipe Barrientos (Pontificia Universidad Católica de Chile)

Bayesian density estimation for compositional data using random Bernstein polynomials We propose a Bayesian nonparametric model for single density estimation, for data in the p-dimensional simplex space, say S_p. The proposal is based on a particular class of multivariate Bernstein polynomials on S_p and extends the Dirichlet-Bernstein prior for density estimation, for data in a closed, bounded interval. The resulting model corresponds to expressing the density of the data as a particular mixture of Dirichlet distributions. We show that these mixtures…

Find out more »### Peter Müller (University of Texas at Austin)

A Nonparametric Bayesian Model for Local Clustering We propose a nonparametric Bayesian local clustering (NoB-LoC) approach for heterogeneous data. Using genomics data as an example, the NoB-LoC clusters genes into gene sets and simultaneously creates multiple partitions of samples, one for each gene set. In other words, the sample partitions are nested within the gene sets. Inference is guided by a joint probability model on all random elements. Biologically, the model formalizes the notion that biological samples cluster differently with respect to different genetic processes, and that each…

Find out more »## May 2013

### Stephan Poppe (University of Leipzig)

Species Sampling Processes: predicting the unpredictable and estimating measures of diversity The sampling of species problem relates to the issue of how to infer the relative species abundances from finite data, when many species occurring in the population are not present in the sample. Although these abundances can be seen to be the ultimate measure of the diversity in a population, there is also some interest in estimating particular summarizing diversity indexes such as the Shannon index and the actual number…

Find out more »### Alessio Farcomeni (University of Rome La Sapienza)

Semiparametric capture-recapture with heterogeneous capture probabilities Capture-recapture experiments are commonly used to estimate the size of a closed population. Link (2003) has underlined identifiability problems when one wants to make inference with heterogeneous capture probabilities in a semiparametric framework. If subject-specific capture probabilities are random effects with no assumption on the mixing distribution, the conditional likelihood is not identifiable. Link (2003) invokes the equivalence of conditional and complete likelihood (Sanathanan, 1972) to conclude that semiparametric inference is not possible in…

Find out more »## April 2013

### Fan Li (Duke University)

Bayesian inference for regression discontinuity designs with application to Italian university grants evaluations Regression discontinuity (RD) designs are usually interpreted as local randomized experiments: A RD design can be considered as though it were a randomized experiment for units with a realized value of a so-called forcing variable falling immediately around a pre-fixed threshold. Motivated from the evaluation of Italian university grants, we consider a fuzzy RD where the receipt of the treatment is based on eligibility criteria and a…

Find out more »## March 2013

### Benedicte Haas (Université Paris-Dauphine)

On scaling limits of Markov branching trees Probabilists and combinatorists are interested since a long time in the asymptotic description of large random trees, as, for example, large uniform trees (chosen uniformly at random in a certain class of trees) or large conditioned Galton-Watson trees. After recalling classical results on that topic, we will develop the case of a family of random trees satisfying a Markov branching property. Roughly, this property says that the subtrees above some given height are…

Find out more »### Omar El-Dakkak (Université Paris Ouest)

Exchangeable Hoeffding decompositions: characterizations and counterexamples Since the pioneering work of Hoeffding in 1948, the so-called Hoeffding-ANOVA decompositions proved to be a very effective tool in obtaining limit theorems and have been widely used in various applications. In this talk, we present the main elements of the theory of Hoeffding decompositions for (infinitely extendible) exchangeable sequences, as it has developed in recent years. We start by presenting a necessary and sufficient condition, due to G. Peccati, for an exchangeable sequence…

Find out more »### Giovanni Peccati (University of Luxembourg)

Universality and chaos I will describe some recent advances involving universality results for homogeneous sums, both in a classic and free setting. Many connection with influence functions, as well as applications to random matrices will be highlighted.

Find out more »## February 2013

### Dan Roy (University of Cambridge)

The combinatorial structure underlying the beta process is that of a continuum of Blackwell-MacQueen urn schemes We uncover a novel urn scheme underlying conditionally independent sequences of Bernoulli processes that share a common beta process hazard measure. As shown by Thibaux and Jordan (2007), in the special case when the underlying beta process has a constant concentration function and a finite and non-atomic base measure, the combinatorial structure is that of the Indian buffet process (IBP) introduced by Griffiths and…

Find out more »## January 2013

### Michael Möller (Dortmund University)

Integrals and cubature sums

Find out more »## December 2012

### Silvia Montagna (Duke University)

Computer emulation with non-stationary Gaussian processes Computer codes are used widely in modern scientific research in complex chemical, thermodynamical and astrophysical processes. These codes deterministically map vectors of high-dimensional inputs into a scalar or vector-valued output, and must be run for many different input configurations to provide an adequate knowledge of the response surface. However, these computer models are very expensive to evaluate for all input values of interest. Therefore, there is often interest in building a statistical model also…

Find out more »## November 2012

### Nicola Sartori (University of Padova)

Calibrating hybrid pseudo likelihood ratios for a parameter of interest For inference about a parameter of interest in the presence of nuisance parameters, we consider a pseudo likelihood obtained from a genuine or composite likelihood by replacing the nuisance component with an estimate based on a generic estimating equation. Suitable adjustments are developed for the resulting pseudo likelihood ratio statistic, taking into account both nuisance estimation procedure and possible misspecification. Joint work with L. Pace and A. Salvan

Find out more »### Alessandra Luati (University of Bologna)

The generalised autocovariance function The generalised autocovariance function is defined for a stationary stochastic process as the inverse Fourier transform of the power transformation of the spectral density function. Depending on the value of the transformation parameter, this function nests the inverse and the traditional autocovariance functions. A frequency domain non-parametric estimator based on the power transformation of the pooled periodogram is considered and its asymptotic distribution is derived. The results are employed to construct classes of tests of the…

Find out more »### Ron S. Kenett (KPA Ltd., Israel)

Applications of Bayesian Networks to Operational Risks, Healthcare, Biotechnology and Customer Surveys Modelling cause and effect relationships has been a major challenge for statisticians in a wide range of application areas. Bayesian Networks combine graphical analysis with Bayesian analysis to represent descriptive causality maps linking measured and target variables. Such maps can be used for diagnostics and predictive analytics. The talk will present an introduction to Bayesian Networks and their applications to web site usability (Harel et al, Kenett et…

Find out more »### Botond Szabo (Eindhoven University of Technology)

On frequentist coverage of Bayesian credible sets Adaptive techniques for nonparametric estimation have been widely stud- ied in the literature and many rate-adaptive results have been provided for a variety of statistical problems. However an adaptive estimator without any knowledge of its uncertainty is rather uninformative, since one knows that the estimator is optimally close to the true function, but has no information about the actual distance. In the Bayesian framework credible sets can be constructed to quantify the uncertainty…

Find out more »## September 2012

### Matteo Marsili (Abdus Salam International Center for Theoretical Physics, Italy)

Collaboration in social networks The very notion of social network implies that linked individuals interact repeatedly with each other. This notion allows them not only to learn successful strategies and adapt to them, but also to condition their own behavior on the behavior of others, in a strategic forward looking manner. Game theory of repeated games shows that these circumstances are conducive to the emergence of collaboration in simple games of two players. We investigate the extension of this concept…

Find out more »## June 2012

### Fabrizia Mealli (University of Florence)

Using secondary outcomes and covariates to sharpen inference in randomized experiments with noncompliance Restrictions implied by the randomization of treatment assignment on the joint distribution of a primary outcome and an auxiliary variable are used to tighten nonparametric bounds for intention-to-treat effects on the primary outcome for some latent subpopulations, without requiring the exclusion restriction assumption of the assignment. The auxiliary variable can be a secondary outcome or a covariate, while the subpopulations are defined by the values of the…

Find out more »## May 2012

### Natalia Bochkina (University of Edinburgh)

The Bernstein - von Mises theorem: relaxing its assumptions and extending it to nonregular models The Bernstein - von Mises theorem is an important result in Bayesian asymptotics, giving conditions under which the posterior distribution of a finite-dimensional parameter can be approximated by the Gaussian distribution. On one hand, this result quantifies consistency and efficiency of Bayesian procedures whichmakes them ``optimal'' from the frequentist point of view, and, on the other hand, it justifies Gaussian approximation of the posterior distribution…

Find out more »### Alessandra Giovagnoli (University of Bologna)

Design of experiments: from physical to simulated Since Fisher’s times, the principles for planning scientific experiments correctly have been at the heart of the statistical debate. This is particularly important in a clinical context, for ethical as well as inferential reasons. After a brief excursus through the history of experimental design, this presentation will deal with recent developments in the design of clinical trials ethically aimed at minimizing the negative impact on patients or healthy volunteers. It will include issues…

Find out more »## April 2012

### David Knowles (University of Cambridge)

Diffusion trees as priors The Dirichlet diffusion tree has attractive theoretical properties and empirical performance on various tasks. We present an extension which removes the restriction to binary trees allowing arbitrary branching structure, the Pitman Yor diffusion tree. We show this process is exchangeable and projective. Both the DDT and PYDT can be constructed as continuum limits of nested CRP models. We demonstrate efficient deterministic and sampling based inference. Finally we discuss some open modelling questions, such as whether to…

Find out more »## March 2012

### Matthias Birkner (Johannes Gutenberg University Mainz, Germany)

Ancestral lineages under local regulation The spatial embeddings of genealogies in models with fluctuating population sizes and local regulation are relatively complicated random walks in a space-time dependent random environment. They seem presently not well understood. We use the supercritical discrete-time contact process on Z^d as the simplest non-trivial example of a locally regulated population model and study the dynamics of an ancestral lineage sampled at stationarity, viz. a directed random walk on a supercritical directed percolation cluster. We prove…

Find out more »## February 2012

### Alessandro Arlotto (University of Pennsylvania)

"Optimal Hiring and Retention Policies for Heterogeneous Workers who Learn"

Find out more »### Antonio Colangelo (European Central Bank)

Banks' Balance Sheet Statistics and Financial Flows in the Euro Area The analysis of money and credit developments is core to the conduct of monetary policy. Those statistics are constructed in the euro area starting from the balance sheets of resident banks. By appropriately netting banks' cross-positions in the euro area, aggregated positions are identified with euro area resident non-banks, leading to the main credit and monetary aggregates, as well as with non-euro area residents. Financial innovation (e.g. securitisation, activities…

Find out more »## November 2011

### Yee Whye Teh (University College London)

Efficient MCMC for Continuous Time Discrete State Systems A variety of phenomena are best described using dynamical models whichoperate on a discrete state space and in continuous time. Examplesinclude Markov jump processes, continuous time Bayesian networks,renewal processes and other point processes, with applications rangingfrom systems biology, neuroscience, genetics, computing networks andhuman-computer interactions. Posterior computations typically involveapproximations like time discretization and can be computationallyintensive. In this talk I will describe our recent work on a class ofMarkov chain Monte Carlo methods…

Find out more »### Andrés Christen (CIMAT, México)

Towards Uncertainty Quantification and Inference in the stochastic SIR Epidemic Model We introduce a novel method to conduct inference with models defined through a continuous-time Markov process, and we apply these results to a classical stochastic SIR model as a case study. We obtain approximations for first and second moments for the state variables. These approximate moments are in turn matched to the moments of an inputed generic discrete distribution aimed at generating an approximate likelihood that is valid both…

Find out more »## October 2011

### Ilya Molchanov (University of Bern)

Symmetries of probability distributions, their geometric meaning and financial applications The talk starts with the known put-call symmetry property and its application to semi-static hedging of barrier options. Then it is explained how to interpret this property geometrically and extend it in various ways, most importantly for the multivariate (multiasset) case that would correspond to basket and exchange options in financial language. In particular, the symmetry property for the exchange case is a weakening of the famous exchangeability property of…

Find out more »## September 2011

### Sergio Bacallado (Stanford University)

A Bayesian analysis of reversible time series with an uncertain length of memory We propose a Bayesian analysis of reversible time series using a Probabilistic Suffix Automaton (PSA) model. We show that PSAs have a representation as higher-order Markov chains, and that the class of reversible PSAs generalize reversible variable-order Markov chains. The analysis uses a conjugate prior for higher-order Markov chains (Bacallado, Annals of Statistics, 39 (2), 2011), which allows us to sample the posterior of the process and…

Find out more »## May 2011

### Keisuke Hirano (University Of Arizona)

"Impossibility Results for Nondifferentiable Functionals"

Find out more »### Eugenio Regazzini (University of Pavia)

"Finitely additive probabilities in statistics"

Find out more »### Hykel Hosni (Scuola Normale di Pisa)

"Subjective probability as a normative theory of rational belief"

Find out more »## April 2011

### Alexander V. Gnedin (Utrecht University)

"Extensions of Mallows' distribution on permutations and the q-exchangeability"

Find out more »## March 2011

### Peter Orbanz (University of Cambridge)

"Projective limit techniques in Bayesian nonparametrics"

Find out more »### Sinead Williamson (University of Cambridge)

"Dependent completely random measures via Poisson line processes"

Find out more »## December 2010

### Michele Guindani (MD Anderson Cancer Center, Houston)

"Generalized Species Sampling Priors with Latent Beta reinforcements"

Find out more »### Henry Wynn (London School of Economics)

"Algebraic Statistics: some background and advances"

Find out more »## November 2010

### Robert C. Griffiths (University of Oxford)

"Stochastic processes with polynomial eigenfunctions"

Find out more »