Local asymptotic normality for an ergodic diffusion with jumps

In a recent paper Eulalia Nualart with co-authors consider a multidimensional ergodic diffusion with jumps driven by a Brownian motion and a Poisson random measure associated with a compound Poisson process, whose drift coefficient depends on an unknown parameter. They show that the local asymptotic normality property holds in this case. The paper has just been accepted for publication in Statistics: A Journal of Theoretical and Applied Statistics. The arXiv version can be found here.

“Maximum likelihood estimation for linear Gaussian covariance models” to appear in Journal of Royal Statistical Society – Series B

In a new paper  Piotr Zwiernik, Caroline Uhler, and Donald Richards study the Gaussian models with linear structure of the covariance matrix. Maximum likelihood estimation for this class of models leads to a non-convex optimization problem which typically has many local maxima. Using recent results on the asymptotic distribution of extreme eigenvalues of the Wishart distribution, they provide sufficient conditions for any hill-climbing method to converge to the global maximum. An important consequence of this analysis is that for sample sizes n14p, maximum likelihood estimation for linear Gaussian covariance models behaves as if it were a convex optimization problem. The paper is going to appear in Journal of Royal Statistical Society – Series B; the arXiv version is available here.

David Rossell joining the statistics group

01 David Rossell is a recipient of this year’s Ramon y Cajal fellowship and he will join our group in September as a visiting professor. Prof. Rossell is on leave from the University of Warwick where he is an Associate Professor in the Statistics Department. His research interests include experimental design, model building and selection, and dimensionality reduction, all with emphasis on high-throughput Bioinformatics experiments and the Bayesian approach.

BGSMath Monthly Programs 2017

Gràfica-Monthly-ProgramsGábor Lugosi and Piotr Zwiernik are involved in two research programs that have just received funding from BGSMath in the framework of the “María de Maeztu“ grant. BGSMath will sponsor in total three programs in 2017. Gábor will co-organize a 5-week program “Random discrete structures and beyond” and Piotr will co-organize a 4-week program “Algebraic and combinatorial phylogenetics“. More details will follow soon.

Florian Simatos on queue-based CSMA protocols

risingstar-2014In the next Statistics and Operation Research Seminar Florian Simatos will talk about “Delay performance of queue-based CSMA protocols”.

When: Thursday, June 9, at 12:00pm     Where: 24.021.

Dr Simatos obtained his PhD at INRIA (France) in 2009 and held research positions at the Center for Mathematics and Computer Science (CWI) and at Eindhoven University of Technology (TU/e). He is the recipient of the 2014 ACM SIGMETRICS Rising Star Researcher Award. His research lies at the boundary of the performance evaluation and applied probability.


Lorenzo Rosasco on optimal learning with stochastic projection regularization

rosasco_azNext week Lorenzo Rosasco (Genoa/MIT) will give a talk on “Less is more: optimal learning with stochastic projection regularization”, which aims at presenting results from the following two papers:

When: Thursday, June 2, at 12pm. Where: 24.021.

The research of Dr. Rosasco focuses on studying theory and algorithms for machine learning. He has developed and analyzed methods to learn from small as well as large samples of high dimensional data, using analytical and probabilistic tools, within a multidisciplinary approach drawing concepts and techniques primarily from computer science but also from statistics, engineering and applied mathematics.


“Total positivity in Markov structures” to appear in the Annals of Statistics

Totally positive distributions appeared independently in many context in statistics and they make a far generalization of a bivariate distribution with a positive covariance between the components. What seems to be a severe restrictions, turns out to be a natural assumption when additional conditional independence is present.

In this paper Piotr Zwiernik and co-authors study positive dependence in the context of conditional independence and conditional independence models. This is a beginning of a greater program of understanding total positivity in the context of graphical models, which links to high-dimensional inference procedures.