Department of Statistics and
Applied Probability
th
10 Anniversary Symposium
Date: 25th February 2008 (Monday)
Venue: University Hall Auditorium, Level 2, Lee Kong
Chian Wing
Programme
Time
Programme
8.45 - 9.15am
Registration
9.00 - 9.15am
Guests to be seated
9.15 - 10.15am
Opening addresses and presentation on history and development of DSAP
10.15 - 10.30am
Photo Session
10.30 - 11.00am
Tea Break
Hemodynamic response function using event-based fMRI
Young Truong
University of North Carolina at Chapel Hill & Former Head of DSAP
11.00 - 11.30am
Abstract
Event-related fMRI (ER-fMRI) has played an important role in many recent brain imaging studies
to explore the relationship between recorded fMRI signals and neural activity. Different from
traditional block-design fMRI, ER-fMRI is very good at estimating the timing and waveform of the
hemodynamic response. Various methods have been proposed in the literature to model the
hemodynamic response function (HRF). Most of them however have a number of limitations. In
this paper, we propose a novel regression approach to estimate the HRF directly. The approach
is based on point process modeling to account for the event-related designs. Compared to the
existing methods, the proposed procedure yields simultaneously the nonparametric estimate of
the HRF and a test for the linearity assumption. To illustrate its usefulness and the scientific
implications, we applied this procedure to study the spatial variation of the HRF, and the extent to
which the linear relationship holds in various regions of interest for Parkinson's disease patients.
This is a joint work with Bai and Huang.
Detection of spatial clustering with generalized likelihood ratio scan test statistics
Chan Hock Peng, DSAP
11.30 - 12.00pm
Abstract
We consider in this talk the detection of spatial clustering in a case-control dataset by using
generalized likelihood ratio (GLR) test statistics to check for significantly large proportion of cases
within some scanning window. In addition to the usual scan test statistic that takes the supremum
GLR value over all windows, we also consider a modified scan test statistic which takes the
average GLR values over all windows. The modification increases the power of the scan statistic
when there are more than one sources of spatial clustering and it does not suffer from the power
loss seen in global clustering test statistics when there is exactly one source of spatial clustering.
Moreover, its log transformation has a simple chi-squared tail probability approximation obtained
by using change of measure and random walk theory. A popular epidemiological dataset is
studied and in contrast to earlier kernel based studies, significant clustering is concluded.
12.00 - 1.30pm
Lunch
A Statistical Model for The Transmission of Infectious Diseases
Xia Yingcun, DSAP
1.30 - 2.00pm
Abstract
The compartmental susceptible-infected-recovered (SIR) model is commonly used in analyzing
the transmission dynamics of infectious diseases. However, in practice it is not easy to use the
model to fit and predict infections due to the unavailability of data required in the model. In this
paper, we propose a statistical time series model, called cumulative alert infection model (CAIM),
based on the transmission mechanism. The model can be estimated by the reported numbers of
incidents alone and allows us to make an instantaneous analysis and suggestions for the
prediction and control of the diseases. The model can incorporate other factors to investigate
relevant issues of infectious diseases, such as immunity and control efficiency. The model is
applied to a number of infection outbreaks, including measles in UK, foot-and-mouth diseases in
UK, and the severe acute respiratory syndrome (SARS) outbreaks in several regions around the
world.
Combining statistical & deterministic approaches in environmental modeling
David Nott, DSAP
2.00 - 2.30pm
2.30 - 3.00pm
Abstract
Environmental modelling often involves the description of complex space-time
processes. Scientific knowledge about these processes is often embodied in complex
deterministic computer models. However, these models are subject to many uncertainties, such
as unknown parameters, uncertain inputs, uncertainties due to limited computational resources
and uncertainties caused by the inadequacy of the deterministic model to capture relevant
physical processes. For these reasons there has been much recent interest in quantifying
uncertainty in deterministic environmental models and more generally in reconciling deterministic
and statistical approaches. This talk will summarize recent developments in this area and
illustrate
with
some
applications
to
rainfall-runoff
modelling
in
hydrology.
Tea Break
Spectrum-based de novo repeat detection in genomic sequences
Choi Kwok Pui, DSAP (joint work with Huy Hoang Do, Franco Preparata, Wing Kin Sung
and Louxin Zhang)
3.00 - 3.30pm
Abstract
A novel approach to the detection of genomic repeats is presented in this paper. The technique,
dubbed SAGRI (Spectrum Assisted Genomic Repeat Identifier), is based on the spectrum (set
of k-mers, for some chosen k) of the genomic sequence. Specifically, the genome is scanned
twice. The first scan (FindHit) detects candidate pairs of repeat-segments, by effectively
reconstructing portions of the Euler path of the (k-1)-mer graph of the genome only
in correspondence with likely repeat sites. This process produces candidate repeat pairs, for
which the location of the leftmost term is unknown. Candidate pairs are then subjected to
validation in a second scan, in which the genome is labelled for hits in the (much smaller)
spectrum of the repeat candidates: high hit density is taken as evidence of the location of the first
segment of a repeat, and the pair of segments is then certified by pairwise alignment. The design
parameters of the technique are selected on the basis of a careful probabilistic analysis (based
on random sequences). SAGRI is compared with three leading repeat-finding tools on both
synthetic and natural DNA sequences, and found to be uniformly and significantly superior in
versatility (ability to detect repeats of different lengths) and accuracy (the central goal of repeat
finding), while being quite competitive in speed.
Optimal Investment and consumption under transaction costs
Lim Tiong Wee, DSAP
3.30 - 4.00pm
4.00 - 4.30pm
Abstract
The problem of optimal investment and consumption was initially studied in an idealized setting
where an investor incurs no transaction costs from trading in a market consisting of a risk-free
asset (``bond'') with constant rate of return and a risky asset (``stock'') whose price is a geometric
Brownian motion with constant rate of return and volatility. Merton (1969, 1971) showed that, for
an investor acting as a price-taker and seeking to maximize expected utility of consumption, the
optimal strategy is to invest a constant proportion (the ``Merton proportion'') of wealth in the stock
and to consume at a rate proportional to wealth. However, the Merton strategy requires
continuous trading and results in an infinite turnover of stock in any finite time interval. In the
presence of transaction costs proportional to the amount of trading, such a continuous strategy is
prohibitively expensive. Davis and Norman (1990) provided a precise formulation of this problem
as one of singular stochastic control and demonstrated how the end-points of the ``notransaction'' region can be obtained using the principle of smooth fit when the horizon is infinite.
In contrast, the finite-horizon investment-consumption problem is much more complex. We
develop a new method for solving the latter problem for which the consumption decision is a
continuous control, and provide algorithms to compute the time-varying buy-sell boundaries and
the value function. Numerical results are also given to illustrate the method.
Tea Break
Estimation of a covariance matrix with zeroes
Sanjay Chaudhuri, DSAP
4.30 - 5.00pm
Abstract
Zeroes in the covariance matrix of a multivariate random vector indicates linear independence
between the corresponding components. Estimation of such covariance matrices have been
discussed, among others, by Anderson (1973), Kauerman (1995), Drton and Richardson (2003),
Cox and Wermuth (2006), Chaudhuri, Drton and Richardson (2007). Due to the presence of
structural zeroes, the maximum likelihood estimator cannot be expressed analytically and can
only be computed iteratively when the underlying distribution is Gaussian. The same algorithm is
used to compute an estimator even if the underlying distribution is non-Gaussian. In this talk we
present an empirical likelihood based estimate of covariance matrices, which preserves the
structural zeroes. This estimator is unique, consistent and convenient to apply in a variety of
situations. Moreover we shall show that it is more efficient than the other estimators for nonGaussian cases and slightly less efficient than the mle for the Gaussian case. Some
generalizations of the methodology will also be discussed.
Boundary proximity of stochastic Loewner evolution
Zhou Wang, DSAP (joint work with Oded Schramm)
Abstract
This paper examines how close the chordal SLE κ curve gets to the real line asymptotically far
away from its starting point. In particular, when κ ∈ (0,4), it is shown that if β > β κ =1/(8/ κ -2),
5.00 - 5.30pm
then the intersection of the SLE κ curve with the graph of the function y=x/(log x) β , x>e, is a.s.
bounded, while it is a.s. unbounded if β = β κ . The critical SLE 4 curve a.s. intersects the graph
α
of y=x − (log log x ) , x>e e , in an unbounded set if α ≤ 1, but not if α >1. Under a very mild
regularity assumption on the function y(x), we give a necessary and sufficient integrability
condition for the intersection of the SLE κ path with the graph of y to be unbounded. We also
prove that the Hausdorff dimension of the intersection set of the SLE κ curve and real axis is 28/ κ when 4< κ <8.
Profile of Faculty Members
BAI Zhidong
Ph.D. (University of Science and Technology of
China)
KUK Yung Cheung, Anthony (Head)
Ph.D. (Stanford University)
Large Dimensional Random Matrices, Urn Models
and Clinical Trials
Bruce BROWN (Visiting Professor)
Ph.D. (Purdue University)
Analysis of Clustered and Longitudinal Data,
Monte Carlo Methods of Inference, Survival
Analysis
LENG Chenlei
Ph.D. (University of Wisconsin-Madison)
Nonparametric Statistics, Categorical Choice Models,
Statistics in Finance
Biman CHAKRABORTY
Ph.D. (Indian Statistical Institute)
Nonparametrics, Survival Analysis, Statistical
Learning
LI Jialiang
Ph.D. (University of Wisconsin-Madison)
Multivariate Analysis, Statistical Computing
Semiparametric Models, Diagnostic Medicine,
Statistical Methods for Epidemiology
LIM Tiong Wee
Ph.D. (Stanford University)
CHAN Hock Peng
Ph.D. (Stanford University)
Change-Point Detection, Sequential Testing
CHAN Yiu Man
Ph.D. (University of Toronto)
Application of Math/Probability to Finance,
Stochastic Control
LOH Wei Liem
Ph.D. (Stanford University)
Forecasting, Business Statistics, Industrial Statistics
Sanjay CHAUDHURI
Ph.D. (University of Washington)
Statistics, Applied Probability
David John NOTT
Ph.D. (University of Queensland)
Graphical Markov Models, Multivariate Statistics,
Decision Theory
CHEN Hsiao Yun, Louis
Ph.D. (Stanford University)
Bayesian Model Selection, Hierarchical Models,
Spatio-temporal Modelling
TEO Yik Ying (Visiting Fellow)
Ph.D. (University of Oxford)
Applied Probability, Stein’s Method, Applications to
Computational Biology
Statistical Genetics, Genomewide Association
Studies
Berwin Ashoka TURLACH
Ph.D. (Université Catholique de Louvain)
CHEN Ying
Ph.D. (Humboldt University in Berlin)
Risk management, Adaptive estimation, Application
of heavy-tailed distribution in financial analysis
CHEN Zehua
Ph.D. (University of Wisconsin-Madison)
Nonparametric Smoothing Techniques,
Statistical Computing, Applications of Statistics
General Statistical Methodology, Statistical Genetics,
Microarray Analysis
CHEW Soon Huat, David (Research Fellow)
Ph.D. (National University of Singapore)
Mathematical Modelling, Bayesian Statistics,
Statistical Filtering Theory
XIA Yingcun
Ph.D. (The University of Hong Kong)
Computational Biology
Semiparametrics, Nonlinear Time Series,
Epidemic Modelling
XU Jinfeng
Ph.D. (Columbia University)
CHOI Kwok Pui (Deputy Head)
Ph.D. (University of Illinois)
Computational Biology, Probability
WU Zhengxiao
Ph.D. (University of Wisconsin-Madison)
Survival Analysis, Statistical Genetics,
Econometrics
CHUA Tin Chiu (Deputy Head)
Ph.D. (Iowa State University)
YAP Von Bing
Ph.D. (University of California, Berkeley)
Survey Sampling, Industrial Statistics, Business
Statistics
GAN Fah Fatt
Ph.D. (Iowa State University)
Computational Biology, Molecular Evolution,
Phylogenetics
Yannis YATRACOS
Ph.D. (University of California, Berkeley)
Statistical Process Control
Parametric and Nonparametric Estimation
Methods, Projection Pursuit Exploratory Data
Analysis, Estimation with Artificial Data
Augmentation
ZHANG Jin-Ting
Ph.D. (The University of North Carolina at
Chapel Hill)
Apratim GUHA
Ph.D. (University of California, Berkeley)
Time Series Analysis and Modelling, Neuroscience
HO Man Wai
Ph.D. (The Hong Kong University of Science and
Technology
Bayesian Nonparametrics, Density and Hazard Rate
Estimation
Nonparametric Regression and Density
Estimation, Functional Data Analysis,
Longitudinal Data Analysis
ZHOU Wang
Ph.D. (The Hong Kong University of Science
and Technology
Brownian Motion, Stochastic Loewner Evolution
Saw Swee Hock Professorship of Statistics
Established in 2003 by former NUS professor and recipient of the inaugural NUS Distinguished
Alumni Service Award, Dr. Saw Swee Hock, the endowed professorship named after him aims to
attract world renowned statisticians to the department to teach courses of their specialty and to
interact with faculty members and students.
2005/06 Saw Swee Hock Professor of Statistics
Professor David Siegmund
Stanford University
2006/07 Saw Swee Hock Professor of Statistics
Professor John David Kalbfleisch
University of Michigan
Email Contact List of Ex-Colleagues
Name/Affiliation
Email
Name/Affiliation
Email
Ayesha Ali
University of Guelph
[email protected]
Lewin-Koh Sock Cheng
[email protected]
Javier Cabrera
Rutgers, The State
University of New Jersey
[email protected]
Liang Faming
Texas A&M University
[email protected]
Chan Kin Yee
SEARS
[email protected]
Lin Ting Kwong
Singapore Management
University
[email protected]
Chan Wai Sum
The Chinese University of
Hong Kong
[email protected]
Lim Keng Suan
National Institute of
Education
[email protected]
Chen Song Xi
Iowa State University
[email protected]
Marriott, Paul Kenneth
University of Waterloo
[email protected]
Cheung Siu Hung
The Chinese University of
Hong Kong
[email protected]
Prabir Burman
University of California,
Davis
[email protected]
Chiang Kok Leong, Andy
Department of Statistics,
Singapore
[email protected] Shao Qiman
University of Oregon &
The Hong Kong University
of Science and Technology
[email protected]
[email protected]
Fan Shenghua, Kelly
California State University,
East Bay
[email protected]
Sutaip Saw Leng Chooi
Nanyang Technological
University
[email protected]
Hsing Tailen
University of Michigan
[email protected]
Stephenson, Alec Grant
Swinburne University of
Technology
[email protected]
Hu Feifang
University of Virginia
[email protected]
Truong, Young Kinh-Nhue [email protected]
University of North Carolina
at Chapel Hill
Kanchan Mukherjee
Lancaster University
[email protected]
Uditha Balasooriya
Nanyang Technological
University
[email protected]
Kwong Koon Shing
Singapore Management
University
[email protected]
Wang Yougan
CSIRO Mathematical and
Information Sciences
[email protected]
Lee Bee Leng
San Jose State University
[email protected]
Yang Zhenlin
Singapore Management
University
[email protected]
Lee Teng Chee, Ronnie
Citibank Singapore
[email protected]
Yeo Kwee Poo
Lilly-NUS Centre
[email protected]
IMS-DSAP
Workshop on High-dimensional Data Analysis
27 – 29 February 2008
IMS Auditorium
With the advent of high throughput technologies and powerful computing facilities, the face of the discipline
of statistics has changed drastically. According to the executive summary of the NSF Report on the Future
of Statistics (http://www.amstat.org/news/nsf4Aug04.pdf), “among the highest priorities for statistics today is
adapting to meet the needs of data sets that are so large and complex that new ideas are required, not only
to analyze the data, but also to design the experiments and interpret the experimental results”.
It is against this background that the IMS and DSAP are jointly organizing a regional workshop, with
participants from China, Hong Kong, Taiwan, India and Singapore, to promote regional networking and
collaboration. The following three sub-themes are identified:
Day 1: Large dimensional random matrices.
Day 2: Functional data analysis.
Day 3: Sparsity issues and model selection in high dimensional problems.
Organizing Committee
-
Zhidong Bai (National University of Singapore)
Arup Bose (Indian Statistical Institute)
Sanjay Chaudhuri (National University of Singapore)
Anthony Kuk (National University of Singapore, Chair)
Albert Lo (Hong Kong University of Science and Technology)
Confirmed Speakers
-
Zhidong Bai, DSAP, NUS, Singapore
Arup Bose, Indian Statistical Institute, India
Probal Chaudhuri, Indian Statistical Institute, India
Jeng-Min Chiou, Institute of Statistical Science, Academia Sinica, Taiwan
Anil Ghosh, Indian Institute of Technology, India
Tailen Hsing, University of Michigan, USA
Su-Yun Huang, Institute of Statistical Science, Academia Sinica, Taiwan
Chii Ruey Hwang, Institute of Statistical Science, Academia Sinica, Taiwan
Thomas Lee, The Chinese University of Hong Kong, Hong Kong
Chenlei Leng, DSAP, NUS, Singapore
Ker-Chau Li, Institute of Statistical Science, Academia Sinica, Taiwan
Debasis Sengupta, Indian Statistical Institute, India
Young Truong, The University of North Carolina at Chapel Hill, USA
Zhonggen Su, Zhejiang University, China
Hansheng Wang, Peking University, China
Shurong Zheng, Northeast Normal University, China
Jin-Ting Zhang, DSAP, NUS, Singapore
Lixing Zhu, The Hong Kong Baptist University, Hong Kong
For more details, please refer to the workshop website,
http://www.ims.nus.edu.sg/Programs/hidim08/index.htm