➢ Plenary Speakers
Peter Guttorp, Professor of Department of Statistics in University of Washington and Norwegian Computing Center
Presentation Topic: Of What Use Is a Statistician in Climate Research?
Abstract
Many issues in climate research deal with uncertainty, in data, in models, and in predictions. Thus it would seem natural that statisticians play an important role in climate science. However, for reasons that I will outline, this has not been the case. I will discuss two statistical analyses related to climate. The first (due to Dave Thompson) shows that the distribution of hemispheric average temperature is strongly related to carbon dioxide emissions, and the second (which I have been working on) looks at how one can go about comparing regional climate model output to temperature data.
Brief Biography
Peter Guttorp received his PhD from University of California Berkeley in 1980. Since then
he has been professor at the University of Washington. He specializes in
stochastic models in atmospheric sciences, environmental sciences, hydrology
and hematology, has published five technical books and about 125 papers. He
is a past president of the International Environmetric Society, a Fellow of
the American Statistical Association, an elected member of the International
Statistical Institute, and a Doctor of Technology honoris causa at Lund
University.
 Chao Agnes Hsiung, Distinguished Investigator and Director, Division of Biostatistics and Bioinformatics, Institute of Population Health Sciences, National Health Research Institutes, Taiwan
Presentation Topic: Shape-Restricted Bayesian Inference using Bernstein Polynomials with Applications in Biomedical Research
Abstract
Shape-restricted Bayesian inference using Bernstein polynomials as priors enjoys the properties that these priors select only smooth functions, have large support, incorporate shape restrictions easily and are easily generated. This approach to shape-restricted inference has been applied to estimation of survival function, isotonic regression and convex regression, among others. In this talk, I will mainly discuss its use in the study of a genome-wide profiling time-course expression of virus genes and in the study of a two-dimensional shape-restricted regression from a health problem. In the former, many salient features of the expression profile like onset time, maximum value, time to maximum value, area under curve etc. can be obtained immediately and many biological questions can thus be formulated quantitatively and insights can be offered to the virus biology. In the latter, Bayesian shape-restricted regression of two or more variables will be applied to study the age-specific incidence of lung cancer over the period from 2000 to 2007 in Taiwan. (This work was jointly with I-Shou Chang , Li-Chu Chien , and Yuh-Jenn Wu.)
Brief Biography
Chao Agnes Hsiung received her undergraduate degree from National Tsing Hua University in 1972 and her PhD from Columbia University in 1975. After a visiting assistant professor position at Cornell University, she returned Taiwan. She was associate professor and promoted to full professor at the National Central University from 1975 to 1985, and a research fellow of the Institute of Statistical Science, Academia Sinica from 1985 to 1997. Since 1997 she has been the Director of the Division of Biostatistics and Bioinformatics at the National Health Research Institutes, Taiwan. She has been working in the field of survival analysis/event history analysis, her current interests include statistics in genetics and genomics, and models and applications in biomedical sciences. She served as the President of the International Chinese Statistical Association (ICSA) in 2001. She was elected as a Fellow of the Institute of Mathematical Statistics (IMS) in 1994, and an elected member of the International Statistical Institutes (ISI) since 1985. She has received awards including the Outstanding Research Award from National Science Council in 1991-1993 and Outstanding Alumnus Award from National Tsing Hua University in 2009. She has been the program director of the project “Genetic Epidemiological Study of Lung Adenocarcinoma in Taiwan” sponsored by the National Genomic Medicine Research Program since 2002. Her e-mail address is hsiung@nhri.org.tw.
Hans R. Künsch, Professor, ETH Zurich, Department of Mathematics, Seminar für Statistik
Presentation Topic: Interactions Between Deterministic and Stochastic Models in Environmental Sciences
Abstract
Deterministic models are predominant in environmental sciences because they represent our knowledge about physical processes and because their parameters have a clear physical interpretation. Statistical methods enter at different levels. In data assimilation, observations are used to correct forecasts and to estimate initial conditions for the next forecast cycle. Since models usually have substantial biases which cannot be attributed to observational errors, Bayesian methods are often used to estimate this bias in order obtain more realistic uncertainty quantifications. Finally in some cases, it is possible to model uncertainties in the inputs or parameters by including a stochastic component in an otherwise deterministic model. I will discuss these approaches, using examples from climate and aquatic systems.
Brief Biography
Hans R. Künsch received his undergraduate degree in 1975 and his PhD in 1980, both from ETH Zurich. After a postdoc position
at the University of Tokyo he became assistant professor at ETH Zurich and was promoted to associate professor with tenure
in 1989 and to full professor in 1992. He is chairing the department of Mathematics from 2007-2009. He was an associate editor of the Annals of Statistics from 1987-1991 and from 1995-1997. From 1998-2000 he was coeditor of the Annals of Statistics together with Jim Berger, and a member of the IMS council 2003-2006. He is a fellow of the IMS since 2000, elected member of ISI since 2008 and gave
an IMS Medallion lecture in 2002. His interests include spatial statistics and random fields, time series analysis, robust statistics, model selection, sequential Monte Carlo methods and environmental modeling. His email address is kuensch@stat.math.ethz.ch
Shige Peng, Professor,
School of Mathematics, Qilu Institute of Finance, Shandong University, Jinan, China
Presentation Topic: Probability and Distribution Model Uncertainties and Nonlinear Expectations
Abstract
We present our new theory of nonlinear expectation to solve problems of model uncertainties of probability and statistical distributions. These problems are well-known in
pricing and risk measuring in financial markets. We present our new law of large number and central limit theorem under sublinear
expectation. Our results show that its limit distribution is a sublinear one, called G-normal distribution. We present a new type
of Brownian motion, called G-Brownian motion, which is a continuous stochastic process with independent and stationary increments under a sublinear expectation, called G-expectation. The corresponding robust version of Ito's calculus is also very useful for problems of risk measure in finance. This theory can be regarded as a new and nontrivial generalization of Backward Stochastic Differential Equations (BSDE).
Brief Biography
Shige Peng is Professor of Mathematics, the Director of Institute of Finance, and the Director of Institute of Mathematics, Shandong University, China. He is the chief scientist of the National Basic Research Program project "Quantitative Analysis and Computation in Financial Risk Control" 2007-2011. He was elected as member of Chinese Academy of Sciences in 2005. He has received many awards, including the second National Natural Science Prize in 1995 and the Top Scientific and Technology Prize by Shandong Province in 2003 and Tan Kah Kee Science Award 2008. His email address is peng@sdu.edu.cn.
Professor Peng's main contributions are as follows:
- Foundation of Theory of Backward Stochastic Differential Equations (BSDE): Peng’s 1990 paper with E. Pardoux is widely considered as "founder paper", "seminal work" of the theory of BSDE. BSDE becomes a very active in frontiers of mathematics, 5 international colloquia in BSDE have been held.
- Nonlinear Feynman-Kac formula: In 1992 Peng discovered and proved that a large part of systems of semi-linear parabolic and elliptic PDE can be expressed as a solution of a BSDE. When the PDE is linear, it becomes a famous and fundamental result in probability theory and physics: the well-known Feynman-Kac formula.
- Obtained General Stochastic Maximum Principle: The well-known “Pontryagin’s principle” is considered as one of three mile-stones in the modern control theory. In Peng’s 1989 paper, published in SIAM Control & Opim, a general stochastic maximum principle was obtained. This achievement in stochastic control has been considered as one of “two major advances have been made in the last two decades”.
- Introduced a dynamic nonlinear expectation theory: the first notion of dynamic nonlinear expectation in continuous time framework was introduced by Peng in 1997. Peng called it g-expectation. He then introduced an original method to obtain the decomposition theorem of g-supermartingales, corresponding to the well-known Doob-Meyer’s classical result. Recently, based on this powerful result, Peng and his coauthors have proved that most dynamic risk measure within a Wiener probability space is in fact a g-expectation and thus can be generated by a simple function g.
Professor Peng was invited by many internationally well-known institutes to give series of systematic lectures, e.g. Princeton University, Osaka University, Ecole Polytechnique, ETH (Zurich), Institut de Henry Poincare, and Columbia University. Many well-known researchers have made important contributions to this new field of BSDE and many institutes and departments organized systematic courses. In the book Backward Stochastic Differential Equation (LONGMAN, 1997), at the very beginning of the Preface, editors N. El-Karoui and L. Mazliak wrote:"Since the founder paper of Pardoux and Peng concerning general existence and uniqueness results appeared in 1990, Backward Stochastic Differential Equations have become a field of increasing activity and interest due in particular to their connections with stochastic optimization problem where they proved to be a powerful and elegant tool to deal with state constraints."
Recently, a new theory of G-expectations, G-normal distribution and G-Brownian motions has been introduced by Peng. This theory was firstly published in the well-known Abel Symposia, 2005. He then has proved a new law of large numbers and a new central limit theorem under probability model uncertainty, i.e., under a sublinear expectation space, which are beautiful and fundamental results (see arXiv:0711.2834v1). He has contributed many invited talks and series of lectures on this new field, e.g., plenary talks in IMS-China International Conference on Statistics and Probability 2008, June 11-13, 2008, Hangzhou, China, as well as in the 1st PRIMA Congress, July 06-10, 2009, UNSW, Sydney, Australia. He will also delivery a plenary talk in the ICM 2010 Congress, Hyderabad, India.
Qihua Wang, Professor, Academy of Mathematics and Systems Science, Chinese Academy of Sciences
Presentation Topic: Sufficient Dimension Reduction with Missing Response at Random
Abstract
Sufficient dimension reduction methods are useful for handling high-dimensional data. However, it is common in practice that eesponses of some subjects are not observed. In this paper we first show that under some minor conditions, the complete-case analysis yields a $n^{1/2}$-consistent estimator of the central subspace of regression. Generally, the missingness carries additional information about the central subspace. However, the complete-case analysis does not make use of information of the missingness and hence leads to loss of effectiveness. For using the information to increase efficiency, we propose a two-stage procedure, named fusion--refinement (FR) procedure. At the first stage, we obtain a subspace including the central subspace by fusing information on regression and missingness. At the second stage, we refine the obtained subspace to recover the central subspace by complete-case analysis and imputation method, respectively. Sliced inverse regression is used to illustrate the proposed procedures. Simulation studies are conducted to compare them, and a data-driven procedure is suggested to choose from the complete-case analysis and the FR procedures for a purpose of a real application. A real data analysis is used to illustrate our methods.
Brief Biography
Qihua Wang got his PhD at Peking University in 1996.
After a postdoc position at the Institute of Applied Mathematics,
Chinese Academy of Sciences (CAS), he becamed a lecturer at Peking University in 1998.
Since 2003, He has been professor at Academy of Mathematics and Systems Science, CAS.
He specializes in survival analysis, complex data analysis and dimension reduction techniques.
He is an Associate Editor of Science in China, Annals of Institute of
Statistical Mathematics, Journal of Applied Probability and Statistics and Acta Math Apply Sinica.
He won the National Science Fund of China for distinguished young scholar in 2007.
He was named a Chang Jiang Scholar by the Ministry of Education in China in 2009.
Mike West, The Arts & Sciences Professor of Statistical Science at Duke University
Presentation Topic: Recent Advances in Bayesian Modelling for Time Series & Spatio-Temporal Problems
Abstract
Among a number of recent advances in stochastic modelling for time series and spatial problems are models for increasingly high-dimensional time series -- both in traditional time series forecasting problems and in large-scale spatial time series. I will discuss and review some of these developments, focusing on models exploiting conditional independence structures with goals of enabling scale-up in computation (Bayesian MCMC methods) while ensuring adaptability and flexibility in modelling complicated patterns of multivariate dependencies in both temporal and spatio-dynamic contexts. The talk will select from recent work with several current and past students and postdoctoral coauthors, and touch on motivating applications in areas such as finance and atmospheric science.
Brief Biography
Mike West is a member of the faculty in statistical
science at Duke University, where he holds a distinguished chair as the Arts & Sciences Professor of Statistical Science.
West is an international research leader in statistics whose special areas of expertise spans a wide range of areas in Bayesian
statistical modeling and computational statistics, and inter-disciplinary applications in science, biomedicine, finance and
other areas.
Professor West led the development the Institute of Statistics and Decision Sciences at Duke University, now the Department of
Statistical Science, from 1990 until 2001. He has played leading roles in the establishment and on founding boards of several
national research institutes in the USA and UK, including the National Institute of Statistical Science, the NSF Statistical
and Applied Mathematical Sciences Institute in the USA, as well as in professional societies and associations, and currently
sits on the boards of national statistical research institutes in the USA, UK and Japan. West is currently the President of
the International Society for Bayesian Analysis – the world’s premier professional organization in Bayesian statistics, and
chair elect of the Bayesian Statistical Science section of the American Statistical Association – the major professional organization
for Bayesian statistics in the USA.
Professor West’s research and teaching activities are in a number of areas of statistics involving methodology development and
applications of complex stochastic modeling in high-dimensional problems. Core areas of modeling research relate to theory and
methods of multivariate analysis, latent structure, high-dimensional inference and computation, stochastic computational methods,
and time series and dynamic modeling, among others. Since his PhD in 1982, West has published about 150 papers in statistics and
disciplinary journals, as well as three books and several edited volumes. West has advised numerous primary PhD students and
postdoctoral associates, most of whom are now in faculty, industrial or governmental positions involving advanced statistical
research.
West's collaborations and inter-disciplinary statistical research have spanned areas including signal processing, finance,
climatology, and others, with a major focus in several biomedical areas including neurophysiology and clinical biostatistics.
In the last decade, this has included an increased and major focus on genomics and systems biology, with major applied research
interests in basic pathway biology and clinical studies. In these areas, West co-directs a number of university-wide, nationally
funded research programs and holds appointments in several interdisciplinary research centers at Duke.
More details are at www.stat.duke.edu/~mw
|
