Fall 2008
All seminars are 4:00  5:00 p.m., in Cullimore Hall
Room 611 (Math Conference Room) unless noted otherwise.
Refreshments are usually served at 3:30 p.m., and talks start at 4:00 p.m.
If you have any questions about a particular seminar, please contact
the person hosting the speaker.



Thursday 
Dr. Kaifeng Lu, Merck Laboratories, Rahway, NJ Sample Size Calculations for the Constrained Longitudinal Data Analysis Model (abstract) 
Sundar Subramanian 
Thursday 
Dr. Ying Wei, Department of Biostatistics, Columbia University Quantile Regression and its Application in Medical Sciences (abstract) 
Sundar Subramanian 
Thursday 
Dr. Bin Cheng, Department of Biostatistics, Columbia University Statistical Assessment and Sample Size Calculation in QT/QTc Prolongation Studies (abstract) 
Chung Chang 
Thursday 
Nan Kong, Educational Testing Services KDependence Coefficient and LMultivariate Association Coefficient (abstract) 
Chung Chang 
ABSTRACTS
Sample Size Calculations for the Constrained Longitudinal Data Analysis Model: The longitudinal data analysis model proposed by Liang & Zeger (2000) uses baseline as well as postbaseline values as dependent variables, and the baseline means are constrained to be the same across treatment groups due to randomization. We would like to address the issue of sample size calculations for this model. General results on the sample size calculations will be shown. The sensitivity of the sample size requirement to the configuration of the correlation structure and dropout pattern will be assessed for both normal and continuous nonnormal data using simulations. The performance of several adhoc approximations for sample size calculations will also be evaluated. Recommendations under various conditions and implementation using a SAS macro will be discussed. Examples from two phase III clinical trials will be used for illustration. Dr. Kaifeng Lu ~ September 11, 2008 
Quantile Regression and its Application in Medical Sciences: Classical least square regression explores the relationship between a response variable Y and its covariates X solely on estimating the covariate effect on the conditional mean E(YX). In contrast, quantile regression describes the covariates effect on the conditional quantilesQ(YX), such as the median, 0.1th or 0.9th quantile. This way, quantile regression provides a more complete picture of the relationship between Y and X. It can be particularly useful when the upper and lower tails of the conditional quantile functions behave very differently from thecentral trend, and the research interest is more focused on the tails of a distribution. The talk consists of two parts: 1) an introduction to the fundamental theories of quantile regression, including estimation,inference tools and goodnessoffit methods. 2) Illustrative examples of its application in medical science, including exploration of early life risk factors for adult obesity based on a New York cohort, and construction of longitudinal growth chart to incorporate with anindividuals prior path. Dr. Ying Wei ~ September 25, 2008 
Statistical Assessment and Sample Size Calculation in QT/QTc Prolongation Studies: To establish noninferiority in QT/QTc prolongation of a test drug with respect to either a placebo or an active control, a thorough QT/QTc study is recommended by ICH (ICH E14, 2005) which concerns statistical inference on the maximal timematched drug effect. The existing statistical methods for assessing such effects suffer either power loss or parameter restriction. We propose a new asymptotic test with small sample correction based on distribution of maximum of correlated random variables under both a parallel group design and a crossover design. Simulations indicate that our proposed test has adequate powers. For design purpose, the impact on power and sample size calculation for routine QT studies with ECG recording replicates under a parallelgroup design and a crossover design is examined. Formulas for sample size calculations with and without adjustment for covariates were derived under both the parallelgroup design and the crossover design. The results indicate that the approach of replicates may require a smaller sample size for achieving the same power when the correlation coefficient between the recording replicates (or repeated measures) is close to 0 (i.e., these replicate ECGs are almost independent). On the other hand, if the correlation coefficient is close to 1, then there is not much gain regardless whether replicate ECGs are considered. In this paper, an approach to identifying optimal allocation between the number of subjects and the number of replicates per subject is proposed for achieving the maximum power under a fixed budget constraint. The proposed approach can also be applied to minimize the cost for a given power. Dr. Bin Cheng ~ November 13, 2008 
KDependence Coefficient and LMultivariate Association Coefficient: Given a system of multiple random variables, a new measure called the Lmultivariate association coefficient is defined using (conditional) entropy. Unlike traditional correlation measures, the Lmultivariate association coefficient measures the multiassociations or multirelations among the multiple variables in the given system; that is, the Lmultivariate association coefficient measures the degree of the association for the given system. The Lmultivariate association coefficient for the system of two random variables is also called the Lbivariate association coefficient. The association measured by the Lmultivariate association coefficient is a general type of association, not any specific type of a linear or nonlinear association. Unlike the Kdependence coefficient, which is an asymmetrical measure, the Lmultivariate association coefficient is a symmetrical measure. A direct application of the Lmultivariate association coefficient is in variables selection or variables reduction. This paper also explores the relationship between the Lmultivariate association coefficient and the Kdependence coefficient. Nan Kong ~ December 4, 2008 