Department of Statistics Seminar
North Carolina State University

presents

 Dr.  Huixia (Judy) Wang, Dr. Donald Martin,  & Dr. Lexin Li

North Carolina State University

 
A Sampler of NCSU Faculty

  •  Dr. Huixia (Judy) Wang

 Quantile Rank Score Test for Linear Models with a Random Effect  

Abstract:  In this talk, I will introduce a rank score test for linear quantile regression models with a random effect. The developed test makes no distributional assumptions, and it does not require estimating the unknown error density function.  The test is shown to be a very valuable complement to the usual mixed model analysis based on Gaussian likelihood.

 

I will point out two applications of the proposed test. One is in GeneChip microarray studies for detecting differentially expressed genes through analyzing probe level measurements. An enhanced method is proposed to improve the test efficiency at small samples by sharing information across the "interesting" genes. Another application is on analyzing biomedical data where measurements are censored due to a fixed quantification limit.

  • Dr. Donald Martin  

Distributions of patterns and statistics in random sequences

Abstract:  A unified methodology for computing distributions of patterns and statistics in random sequences is discussed.  An auxiliary Markov chain is developed such that events in the original sequence correspond to the auxiliary chain lying in a class of states.  Once this association is made, probabilities are computed through tracking the movements of the auxiliary chain through its states. We highlight the variety of distributions that may be computed in this manner, including those for patterns and statistics in "observation" sequences, and also in hidden "state" sequences modeled by probabilistic graphical models such as hidden Markov models, with a focus on interesting applications.

Both past research and future research directions are given.

  • Dr. Lexin Li  

 Sufficient Dimension Reduction with Application to Microarray Data Analysis

Abstract:  Sufficient dimension reduction can aid the analysis of high-dimensional microarray data by transforming the problems to low dimensional projections. The curse of dimensionality is often alleviated, and the informative data visualization may be enabled. In this talk, we

start with an application of a dimension reduction method, sliced inverse regression, to a microarray survival data analysis. This exercise also introduces new challenges to the methodology of sufficient dimension reduction, including the presence of highly correlated predictors, the small-n-large-p problem, and variable selection in the framework of dimension reduction. We next continue the talk with a discussion of some recently proposed regularized dimension reduction methods to address the above challenges.  Some theoretical properties of the proposed methods will be explored, and the analysis of the microarray data will underlie this line of methodology development.

Friday, September 14, 2007

3:35 - 4:35 pm

301 Riddick Hall

Refreshments will be served in the common area of 301 Riddick at 3:00 pm.