Seminar for Statistics

Search

Research Seminar

Would you like to be notified about these presentations via e-mail? Please subscribe here.

×

Modal title

Modal content

Spring Semester 2019

Date / Time Speaker Title Location
22 March 2019
11:30-12:15 		Judith Rousseau
Ceremade, Dauphine Université Paris
Event Details

Research Seminar in Statistics

Title On Bayesian inference in a family of sparse graphs and multigraphs
Speaker, Affiliation Judith Rousseau, Ceremade, Dauphine Université Paris
Date, Time 22 March 2019, 11:30-12:15
Location HG G 19.1
Abstract Int the first part of this work I will present some properties of the class of graphs based on exchangeable point processes: these include in particular the asymptotic expressions for the numbers of edges, nodes and triangles together with the degree distributions, identifying four regimes: (i) a dense regime, (ii) a sparse, almost dense regime, (iii) a sparse regime with power- law behavior, and (iv) an almost extremely sparse regime. We also propose a class of models within this framework where one can separately control the local, latent structure and the global sparsity/power-law properties of the graph and we derive a central limit theorem for the number of nodes and edges in the graph. In the second part I will recall the Bayesian approach to make inference in such graphs, both for simple and multigraphs proposed by Caron and Fox (2017). I will then study the asymptotic behaviour of the posterior distribution, both under well and mis-specified multigraph models.
On Bayesian inference in a family of sparse graphs and multigraphsread_more
HG G 19.1
12 April 2019
15:15-16:00 		Paulo Rodrigues
Bank of Portugal
Event Details

Research Seminar in Statistics

Title Testing for Episodic Predictability in Stock Returns
Speaker, Affiliation Paulo Rodrigues, Bank of Portugal
Date, Time 12 April 2019, 15:15-16:00
Location HG G 26.3
Abstract Standard tests based on predictive regressions estimated over the full available sample data have tended to nd little evidence of predictability in stock returns. Recent approaches based on the analysis of subsamples of the data have been considered, suggesting that predictability where it occurs might exist only within so-called \pockets of predictability" rather than across the entire sample. However, these methods are prone to the criticism that the sub-sample dates are endogenously determined such that the use of standard critical values appropriate for full sample tests will result in incorrectly sized tests leading to spurious ndings of stock returns predictability. To avoid the problem of endogenously-determined sample splits, we propose new tests derived from sequences of predictability statistics systematically calculated over sub-samples of the data. Speci cally, we will base tests on the maximum of such statistics from sequences of forward and backward recursive, rolling, and double-recursive predictive sub-sample regressions. We develop our approach using the over-identi ed instrumental variable-based predictability test statistics of Breitung and Demetrescu (2015). This approach is based on partial-sum asymptotics and so, unlike many other popular approaches including, for example, those based on Bonferroni corrections, can be readily adapted to implementation over sequences of subsamples. We show that the limiting distributions of our proposed tests are robust to both the degree of persistence and endogeneity of the regressors in the predictive regression, but not to any heteroskedasticity present even if the sub-sample statistics are based on heteroskedasticity-robust standard errors. We therefore develop xed regressor wild bootstrap implementations of the tests which we demonstrate to be rst-order asymptotically valid. Finite sample behaviour against a variety of temporarily predictable processes is considered. An empirical application to US stock returns illustrates the usefulness of the new predictability testing methods we propose.
Testing for Episodic Predictability in Stock Returnsread_more
HG G 26.3
10 May 2019
15:15-16:00 		Jon Wellner
University of Washington
Event Details

Research Seminar in Statistics

Title Multiplier Processes in Statistics: a new multiplier inequality and applications
Speaker, Affiliation Jon Wellner, University of Washington
Date, Time 10 May 2019, 15:15-16:00
Location HG G 19.1
Abstract Multiplier empirical processes have proved to be one of the key unifying themes in modern empirical process theory, with statistical applications including basic symmetriza- tion methods, bootstrap and resampling theory, and analysis of empirical risk minimization procedures. At the heart of the theory of these multiplier processes, a collection of multiplier inequalities provide the basic tools which drive the theoretical developments. In this talk I will review some some of the basic multiplier empirical processes, and explain their importance for a variety of problems in statistics. I will briefly compare sev- eral multiplier inequalities old and new, and then focus on application of a new multiplier inequality, and discuss one particular statistical application concerning convergence rates of least squares estimators (LSE) in regression models with possibly “heavy-tailed” errors. Particular cases involving sparse linear regression, shape restrictions, and finite sampling empirical processes will be mentioned briefly. (This talk is based on the University of Washington Ph.D. work of Qiyang (Roy) Han.)
Multiplier Processes in Statistics: a new multiplier inequality and applicationsread_more
HG G 19.1
22 May 2019
15:15-16:00 		Haakon Bakka
King Abdullah University
Event Details

Research Seminar in Statistics

Title Fundamental ideas of the stochastic partial differential equations approach to defining random effects
Speaker, Affiliation Haakon Bakka, King Abdullah University
Date, Time 22 May 2019, 15:15-16:00
Location HG G 19.1
Abstract In this talk, we will first give a short background on (local) differential operators, sparse discretisations, and GMRFs, and a practical motivation for continuously indexed models. Then we give a brief overview of some PDEs used to model physical processes, and turn these into SPDEs, getting the corresponding spline penalties, spatial, and spatio-temporal models. We outline a proof method for existence and uniqueness of solutions. These SPDEs naturally give the class of Matern models, and extend this class in several directions that are interesting for applications. We apply the approach to two general settings and produce non-separable and non-stationary space-time models.
Fundamental ideas of the stochastic partial differential equations approach to defining random effectsread_more
HG G 19.1
24 May 2019
15:15-16:00 		Bin Yu
UC Berkeley
Event Details

Research Seminar in Statistics

Title Three principles of data science: predictability, computability, and stability (PCS)
Speaker, Affiliation Bin Yu, UC Berkeley
Date, Time 24 May 2019, 15:15-16:00
Location HG G 19.1
Abstract In this talk, I'd like to discuss the intertwining importance and connections of three principles of data science in the title and the PCS workflow that is built on the three principles. The principles will be demonstrated in the context of collaborative projects in genomics for interpretable data results and testable hypothesis generation. If time allows, I will present proposed PCS inference that includes perturbation intervals and PCS hypothesis testing. The PCS inference uses prediction screening and takes into account both data and model perturbations. Finally, a PCS documentation is proposed based on Rmarkdown, iPython, or Jupyter Notebook, with publicly available, reproducible codes and narratives to back up human choices made throughout an analysis. The PCS workflow and documentation are demonstrated in a genomics case study available on Zenodo.
Three principles of data science: predictability, computability, and stability (PCS)read_more
HG G 19.1
13 June 2019
15:15-16:00 		Andreas Buja
University of Pennsylvania
Event Details

Research Seminar in Statistics

Title Models as Approximations -- A Model-Free Theory of Parametric Regression
Speaker, Affiliation Andreas Buja, University of Pennsylvania
Date, Time 13 June 2019, 15:15-16:00
Location HG G 19.1
Abstract In this talk we will rethink what we do when we apply regression to data. We will do so without assuming the correctness of a fitted regression model. We will think of the parameters of the fitted model as statistical functionals, here called "regression functionals," which apply to largely arbitrary (X,Y) distributions. In this view a fitted model is an approximation, not a "data generating process." A natural question is whether such an assumption-lean framework lends itself to a useful statistical theory. Indeed it does: It is possible to (1) define a notion of well-specification for regression functionals that replaces the notion of correct specification of models, (2) create a well-specification diagnostic for regression functionals based on reweighting the data, (3) prove insightful Central Limit Theorems, (4) clear up the misconception that "model bias" generates biased estimates,(5) exhibit standard errors of the plug-in/sandwich type as limiting cases of the pairs- or (X,Y)-bootstrap, and (6) provide theoretical heuristics to indicate that pairs-bootstrap standard errors may generally be more stable than sandwich standard errors.
Models as Approximations -- A Model-Free Theory of Parametric Regressionread_more
HG G 19.1

Note: if you want you can subscribe to the iCal/ics Calender.

Archive: SS 24 AS 23 SS 23 AS 22 SS 22 AS 21 SS 20 AS 19 SS 19 AS 18 SS 18 AS 17 SS 17 AS 16 SS 16 AS 15 SS 15 AS 14 SS 14 AS 13 SS 13 AS 12 SS 12 AS 11 SS 11 AS 10 SS 10 AS 09

JavaScript has been disabled in your browser