ICSA 2009 APPLIED STATISTICS SYMPOSIUM

June 21st-24th, San Francisco, California, USA

Short Courses

Time Schedule: June 21, 2009

Session 1 - 9:00 - 10:15

Break - 10:15 - 10:30

Session 2 - 10:30 - 12:00

LUNCH - 12:00 - 1:30

Session 3 - 1:30 - 2:45

Break - 2:45 - 3:00

Session 4 - 3:00 - 4:30

	Titles	Time & Location	Instructor(s)
1.	Recent Developments in Practical Bayesian Methods for Clinical Trials	9:00-4:30 June 21 Redwoom Room	Professor Peter F. Thall, MD Anderson Cancer Center, TX, USA
2.	Adaptive Designs in Drug Development	9:00-4:30 June 21 Cypress Room (2^nd floor)	Drs. Sue-Jane Wang and Hsien Ming J. Hung, US FDA, Washington DC
3.	Statistical Leaning and Data Mining	9:00-4:30 June 21 Aspen Room	Professor Tao Shi, Ohio State University, OH, USA and Professor Gareth James, Marshall School of Business, USC, CA, USA
4.	Statistical Methods in Bioinformatics	9:00-4:30 June 21 1/4 MES-Ballroom	Professor Jun Liu, Harvard University, MA, USA

***Note***: Refreshments will be served at the back of each classroom during breaks. Lunch will also be served in each of the classrooms in buffet style.

Course 1: “Recent Developments in Practical Bayesian Methods for Clinical Trials"

Instructor: Dr. Peter F. Thall (rex@mdanderson.org), M.D. Anderson Cancer Center

Location: Redwood Room

Abstract: This one-day short course will cover a wide variety of practical Bayesian methods for clinical trial design and conduct. The course will include numerous illustrations from actual clinical trials. Emphasis will be on methods that have been developed in recent years. As time permits, the topics covered will include (1) dose-finding designs for phase I and phase I/II trials based on multivariate outcomes, and patient-specific dosing, (2) monitoring multiple outcomes, monitoring right-censored event time outcomes and accounting for patient heterogeneity in phase II trials, (3) a new method for computing the effective sample size of a parametric prior, (4) adaptive randomization, (5) phase II-III designs, including a new method that accounts for patient heterogeneity, and (6) designs to evaluate and compare multi-stage dynamic treatment regimes. Many of the illustrative applications will include procedures for eliciting and calibrating priors, incorporating historical data, and using computer simulation to establish a design’s frequentist properties. Attendees should have at least a Masters degree in statistics, or equivalent experience, and an understanding of clinical trials and elementary Bayesian concepts.

About the Instructor: Peter F. Thall is the Anise J. Sorrell Professor in the Department of Biostatistics at M.D. Anderson Cancer Center, where he has been a faculty member since 1991. Dr. Thall is an author of over 145 papers and book chapters in the statistical and medical literature. His research interests include Bayesian statistics, medical statistics and clinical trials. He has presented over 120 invited talks at national and international conferences, academic institutions and federal agencies, and he has given 17 short courses on statistical methods for clinical trials. He has served as an associate editor of Journal of the National Cancer Institute and Statistics in Medicine. Currently, he is an associate editor of Clinical Trials and Biometrics, serves on two external advisory boards for program project grants in oncology, is a member of a Special Emphasis Panel/Scientific Review Group for the National Institute of Neurological Diseases and Stroke, and is an American Statistical Association Media Expert.

Course 2: Adaptive Designs in Drug Development

Instructors: Dr. Sue-Jane Wang (suejane.wang@fda.hhs.gov), Dr. Hsien Ming J. Hung (hsienming.hung@fda.hhs.gov), US FDA

Location: Cypress Room (2^nd floor)

Abstract: As the costs increase dramatically, a typical clinical trial carries a high expectation that the trial is able to answer many study questions and subsequently the level of difficulty in conducting the trial rises significantly. Traditional non-adaptive fixed design methodology is therefore often deemed insufficient to achieve the many goals of the trial. The recent advances in adaptive design methodology have been made for evaluation of an experimental treatment, ranging widely from a new look of sample size re-estimation to a mid-term change of statistical decision tree, such as alpha allocation. This short course will give a brief overview of some interesting major advances and present the scenarios where some types of adaptation may be worthy of and needs further exploration. Topics to be covered include: role of adaptive design, learn versus confirm paradigm, sample size re-estimation, adaptive design versus adaptive strategy, adaptive selection of dose, and statistical inference issues with adaptive design, logistics issues.

About the Instructors:   Dr. Sue-Jane Wang, Associate Director for Adaptive Design and Pharmacogenomics, Office of Biostatistics, Office of Translational Sciences, Center for Drug Evaluation and Research, US FDA Dr. H.M. James Hung, Director, Division of Biometrics I, Office of Biostatistics, Office of Translational Sciences, Center for Drug Evaluation and Research, US FDA

Course 3: Statistical Leaning and Data Mining

Instructors: Dr. Tao Shi (Ohio State University) and Dr. Gareth James (University of Southern California)

Location: Aspen Room

Abstract: This one-day short course will cover a collection of methods that have been developed in recent years in statistical machine learning and data mining. This course focuses on both supervised learning (regression and classification) and unsupervised learning (clustering, dimension reduction and data visualization). If time permits, semi-supervised learning will also be introduced. Specifically, the topics will include: validation and cross-validation, ridge regression, lasso and other regularization methods, support vector machines and kernel methods, boosting and random forests; clustering algorithms, spectral methods, linear and nonlinear dimensionality reduction methods, and data visualization tools.

About the Instructors: Tao Shi is an Assistant Professor in the Department of Statistics at the Ohio State University since 2005. His research interests lie in machines learning, statistical methodology and computation on massive data, and statistical applications in atmospheric science and geoscience. Gareth James is an Associate Professor in the Marshall School of Business at the University of Southern California, where he has been a faculty member since 1998. One of his research interests is in the area of statistical learning, including boosting, classification and clustering, high-dimensional data and variable selection. He also works extensively in the area of functional data analysis and is interested in applications to bioinformatics, marketing and finance.

Course 4: Statistical Methods in Bioinformatics

Instructors: Professor Jun Liu <jliu@stat.harvard.edu>, Harvard University

Location: 1/4 Ballroom (Maple-Elm-Sycamore)

Abstract: Over the last decade, Monte Carlo methods have been crucial in many scientific endeavors, ranging from physics to biochemistry, and have become very popular in the statistics community. Many of research problems especially in bioinformatics, which has gained increasing prominence in statistical research, involve high-dimensional Monte Carlo computation. This short course will provide a comprehensive overview of this area and an in-depth introduction of the history, basic ideas, constructions, and applications of various most updated Monte Carlo methods.

About the Instructors:   Professor Liu is the leading scholar in this field and he is the best person possible to present this course. His track record is incredible, including the 2002 COPSS award, ASA and IMS fellows, the 2000 Mitchell Prize for the Best Bayesian Application Paper, and countless other honors in addition to a prolific publication record. Professor Liu has excellent communication skills and has offered various short courses in several professional conferences during last few years, including an ASA CE short course on “Statistical Methods in Bioinformatics,” at the 2004 JSM in Toronto, which was well attended. The topics covered in the four 2-hour sessions address key issues in Markov chain Monte Carlo sampling and the cutting-edge Monte Carlo methodologies with applications in bioinformatics, and strike an excellent balance between theory and application, avoiding excessive technical detail, but covering all the major areas of the Monte Carlo Methods in Bayesian modeling and computation. His Springer book Monte Carlo Strategies in Scientific Computing will be used as the textbook for this short course.