Lecture

David Sprott Distinguished Lecture by Professor Peter Diggle, Lancaster University

Greg Preston — Tue, 14 Mar 2017 13:43:55 +0000

David Sprott Distinguished Lecture by Professor Peter Diggle, Lancaster University Greg Preston Tue, 03/14/2017 - 09:43

A Tale of Two Parasites: how can Gaussian processes contribute to improved public health in Africa?

In this talk, I will rst make some general comments about the role of statistical modelling in scientic research, illustrated by two examples from infectious disease epidemiology. I will then describe in detail how statistical modelling based on Gaussian spatial stochastic processes has been used to construct region-wide risk maps to inform the operation of a multi-national control programme for onchocerciasis (river blindness) in equatorial Africa. Finally, I will describe work-in progress aimed at exploiting recent developments in mobile microscopy to enable more precise local predictions of community-level risk.

About Peter Diggle:

Peter Diggle is a Distinguished University Professor of Statistics in the Faculty of Health and Medicine, Lancaster University. He also holds Adjunct positions at Johns Hopkins, Yale and Columbia Universities, and was president of the Royal Statistical Society between July 2014 and December 2016. Peter began his academic career at the University of Newcastle upon Tyne in 1974, moved to Australia in 1984 as a research scientist with the Commonwealth Scientific and Industrial Research Organisation and returned to the UK in 1988 to take up his current post in Lancaster. His research involves the development of statistical methods for spatial and longitudinal data analysis, and their application to substantive research in the biomedical and health sciences.

David Sprott Distinguished Lecture by Professor David Donoho, Stanford University

Ryan Ball — Tue, 04 Oct 2016 19:13:19 +0000

David Sprott Distinguished Lecture by Professor David Donoho, Stanford University Ryan Ball Tue, 10/04/2016 - 15:13

Factor Models and PCA in light of the spiked covariance model

Principal components analysis and Factor models are two of the classical workhorses of high-dimensional data analysis, used literally thousands of times a day by data analysts the world over. But now that we have entered the big data era, where there are vastly larger numbers of variables/attributes being measured that ever before, the way these workhorses are deployed needs to change.

In the last 15 years there has been tremendous progress in understanding the eigenanalysis of random matrices in the setting of high-dimensional data in particular progress in understanding the so-called spiked covariance model. This progress has many implications for changing how we should use standard `workhorse' methods in high-dimensional settings. In particular it vindicates Charles Stein's seminal insights from the mid 1950's that shrinkage of eigenvalues of covariance matrices is essentially mandatory, even though today such advice is still frequently ignored. We detail new shrinkage methods that flow from random matrix theory and survey the work of several groups of authors.

About Dr. Donoho:

vid L. Donoho is a Professor of Statistics and the Anne T and Robert M Bass Professor of the Humanities and Sciences at Stanford University. He earned his AB in Statistics from Princeton and his PhD in Statistics from Harvard. He began his career in the Department of Statistics at the University of California Berkeley and later moved to the Department of Statistics at Stanford University. He has also worked for Western Geophysical Company and Renaissance Technologies. He was co-founder of network management software company BigFix. His publication list covers Robust Statistics, Signal and Image Processing, Mathematical Statistics, Harmonic Analysis, Scientific Computing, and High Dimensional Geometry. He has made ground-breaking contributions to theoretical and computational statistics, as well as to signal processing and harmonic analysis. His algorithms contributed profoundly to the understanding of the maximum entropy principle, of the structure of robust procedures, and of sparse data description.

He is a member of the United States National Academy of Sciences as well as a foreign associate of the Academie des Sciences of France and has been named a MacArthur Fellow, a Fellow of the American Academy of Art and Sciences, a Fellow of the Society for Industrial and Applied Mathematics, and a Fellow of the American Mathematical Society. He has received the COPSS Presidents’ Award, the John von Neumann Prize, and the Norbert Wiener Prize. He holds an honorary doctorate from the University of Chicago and in 2013 became a Shaw Prize Laureate in the Mathematical Sciences.

David Sprott Distinguished Lecture by Professor David Donoho Poster (PDF)

David Sprott Distinguished Lecture by Martin Wainwright, University of California, Berkeley

Karen Richardson — Thu, 15 Oct 2015 12:47:12 +0000

David Sprott Distinguished Lecture by Martin Wainwright, University of California, Berkeley Karen Richardson Thu, 10/15/2015 - 08:47

Some new phenomena in high-dimensional statistics and optimization

Statistical models in which the ambient dimension is of the same order
or larger than the sample size arise frequently in different areas of
science and engineering. Examples include sparse regression in
genomics; graph selection in social network analysis; and low-rank
matrix estimation in video segmentation. Although high-dimensional
models of this type date back to seminal work of Kolmogorov and
others, they have been the subject of especially intensive study over
the past decade, and have interesting connections to many branches of
applied mathematics and computer science, including random matrices
and algorithms, concentration of measure, convex geometry, and
information theory. In this talk, we discuss various issues in
high-dimensional statistics, including vignettes on phase transitions
in high-dimensional graph recovery, and randomized sketching for
large-scale optimization.

David Sprott distinguished lecture by Raymond J. Carroll, Texas A&M University

Karen Richardson — Wed, 19 Aug 2015 16:54:06 +0000

David Sprott distinguished lecture by Raymond J. Carroll, Texas A&M University Karen Richardson Wed, 08/19/2015 - 12:54

Constrained maximum likelihood estimation for model calibration using summary-level information from external big data sources.

Information from various public and private data sources of extremely large sample sizes are now increasingly available for research purposes. Statistical methods are needed for utilizing information from such big data sources while analyzing data from individual studies that may collect more detailed information required for addressing specific hypotheses of interest. We consider the problem of building regression models based on individual-level data from an "internal'' study while utilizing summary-level information, such as information on parameters for reduced models, from an "external'' big-data source. We identify a set of constraints that link internal and external models. These constraints are used to develop a framework for semiparametric maximum likelihood inference that allows the distribution of the covariates to be estimated using either the internal sample or an external reference sample. We develop extensions for handling complex stratified sampling designs, such as case-control sampling, for the internal study. Asymptotic theory and variance estimators are developed for each case. We use simulation studies and a real data application to assess the performance of the proposed methods.

This is joint work with Nilanjan Chatterjee (Johns Hopkins), Yi-Hau Chen (Academia Sinica, Taipei) and Paige Maas (National Cancer Institute)

David Sprott Distinguished Lecture by Jerome Friedman

Elizabeth Scott — Wed, 25 Feb 2015 18:47:50 +0000

David Sprott Distinguished Lecture by Jerome Friedman Elizabeth Scott Wed, 02/25/2015 - 13:47

Sparsity, boosting and ensemble methods

Statistical or machine learning involves predicting future outcomes from past observations. Many present day applications involve large numbers of predictor variables, sometimes much larger than the number of cases or observations available to train the learning algorithm. In such situations traditional statistical methods fail. Applying regularization techniques can often produce accurate predictions in these settings. This talk will describe the basic principles underlying regularization and then focus on those methods that attempt to exploit sparsity of the predicting model. A fast gradient boosting algorithm is described that can implement a wide variety of regularization methods for linear predictive models. It is then extended to nonlinear modeling giving rise to general learning ensembles.

Jerome Friedman, Stanford University

Jerome Friedman is professor of statistics at Stanford University, where he has held a faculty position since 1982. His outstanding contributions to statistical methods and computer science in the areas of nonparametric statistics and machine learning have led to many honours and awards. His two books Classification and Regression Trees (1984, co-authored with Leo Breiman, Richard Olshen, and Charles Stone) and The Elements of Statistical Learning: Data Mining, Inference and Prediction (2001, co-authored with Trevor Hastie and Rob Tibshirani), are among the most widely used in statistics, machine learning, and data mining.

Everyone welcome.
Reception will follow in the Bruce White Atrium.
Co-sponsored by Department of Statistics and Actuarial Science and Google.

See "Sparsity, Boosting and Ensemble Methods" lecture poster (PDF)

First annual distinguished lecture by David Spiegelhalter

Elizabeth Scott — Tue, 24 Feb 2015 16:57:43 +0000

First annual distinguished lecture by David Spiegelhalter Elizabeth Scott Tue, 02/24/2015 - 11:57

Don't know, can't know: Communicating risk and deeper uncertainty

David Spiegelhalter, Winton Professor of the Public Understanding of Risk, University of Cambridge

Mathematical models are vital whenever we admit we cannot predict precisely what is going to happen, for example in weather forecasting, insurance, nuclear safety, natural disasters, the effect of new medical interventions and, more controversially, in climate change and finance. Such models get so complex that multiple simulations of 'possible futures' may be necessary, which allow us to quantify chances of future events, which then need to be communicated to the public and policy-makers. If we take a Bayesian perspective, then any probability assessment is only a construction based on available information and judgment, and multiple metaphors can be adopted to create a narrative around these quantities.

And models are 'just models', and are always wrong to some extent, and so how can we express this deeper uncertainty?

Reception will follow.
See "Don't know, can't know" lecture poster (PDF)

References:

D.J. Spiegelhalter, I Short, M Pearson. Visualizing uncertainty about the future. Science , 333: 1393-1400, 2011 DOI: 10.1126/science.1191181
D.J. Spiegelhalter, H Riesch. Don't know, can't know: embracing deeper uncertainties when analysing risks. Phil Trans Roy Soc A , 369 4730-4750, 2011. DOI: 10.1098/rsta.2011.0163

David Sprott Distinguished Lecture by Eduardo S. Schwartz

Elizabeth Scott — Tue, 24 Feb 2015 15:56:01 +0000

David Sprott Distinguished Lecture by Eduardo S. Schwartz Elizabeth Scott Tue, 02/24/2015 - 10:56

The real options approach to valuation: challenges and opportunities

This lecture provides an overview of the real options approach to valuation mainly from the point of view of the author who has worked in this area for over 30 years. After a general introduction to the subject, numerical procedures to value real options are discussed. Recent developments on the valuation of complex American options has allowed progress in the solution of many interesting real option problems. Two applications of the real options approach are discussed in more detail: the valuation of natural resource investments, and the valuation of research and development investments.

Eduardo S. Schwartz, University of California

Dr. Schwartz is the California Professor of Real Estate and Professor of Finance, Anderson Graduate School of Management at the University of California, Los Angeles. He has an engineering degree from the University of Chile and a masters and PhD in finance from the University of British Columbia. He has been in the faculty at the University of British Columbia and visiting at the London Business School, the University of California at Berkeley and the Universidad Carlos III in Madrid. His wide-ranging research has focused on different dimensions in asset and securities pricing. Topics in recent years include interest rate models, asset allocation issues, evaluating natural resource investments, pricing Internet companies, the stochastic behavior of commodity prices and valuing patent-protected R&D projects.

David A. Sprott (1930-2013)

Professor David Sprott was the first Chair (1967-1975) of the Department of Statistics and Actuarial Science at the University of ��ݮ��Ƶ and first Dean of the Faculty of Mathematics (1967-1972). The David Sprott Distinguished Lecture Series was created in recognition of his tremendous leadership at a formative time of our department, as well as his highly influential research in statistical science.

Reception will follow in the Mathematics 3 (M3) Bruce White Atrium.

See "The real options approach to valuation: challenges and opportunities" lecture poster (PDF).

David Sprott Distinguished Lecture by Art B. Owen

Elizabeth Scott — Tue, 24 Feb 2015 15:09:54 +0000

David Sprott Distinguished Lecture by Art B. Owen Elizabeth Scott Tue, 02/24/2015 - 10:09

Empirical likelihood

Likelihood methods provide one of the most versatile and effective ways to handle data. They give us tests and confidence intervals with very strong optimality measures. But the cost for using them is usually that we have to know a family of distributions generating our data. Very often we have no reason to assume that any of our usual parametric families truly contain the distribution of our data. Without such a distribution, parametric models can be misleading, i.e., wrong.

Empirical likelihood is a method that provides the benefits of a likelihood function without requiring the user to know a parametric family for the data. The data themselves supply their own parametric family at a quick enough rate to make the resulting tests and confidence intervals reliable. This talk will show some examples of how empirical likelihood can be used.

Art B. Owen, Stanford University

Art Owen is a professor of statistics at Stanford University, where he has held a faculty position since 1985. He obtained a BMath in statistics and computer science from the University of ��ݮ��Ƶ in 1981, and a PhD in statistics from Stanford in 1987. His research interests include empirical likelihood, computer experiments, Monte Carlo and quasi-Monte Carlo sampling, as well as high dimensional transposable style data sets that come up in bioinformatics and Internet applications. His pioneering work on empirical likelihood has opened up a large research area of its own in statistics, and resulted in a very highly cited book on the subject, which he published in 2001.

David A. Sprott (1930-2013)

Reception will follow in the Mathematics 3 (M3) Bruce White Atrium.
See "Empirial Likelihood" presentation slides (PDF).

David Sprott distinguished lecture by William Woodall, Virginia Tech

Elizabeth Scott — Tue, 24 Feb 2015 14:15:55 +0000

David Sprott distinguished lecture by William Woodall, Virginia Tech Elizabeth Scott Tue, 02/24/2015 - 09:15

Monitoring and Improving Surgical Quality

Some statistical issues related to the monitoring of surgical quality will be reviewed in this presentation. The important role of risk-adjustment in healthcare, used to account for variations in the condition of patients, will be described. Some of the methods for monitoring quality over time, including a new one, will be outlined and illustrated with examples. The American College of Surgeons National Surgical Quality Improvement Program (NSQIP) will be described, along with a case study demonstrating significant improvements in surgical infection rates and mortality.