David A. Harville received his Ph.D. degree in 1965 from Cornell University, where he specialized in statistics and animal breeding. Since that time, he has been employed in research and/or graduate education (as a statistician) by the Applied Mathematics Research Laboratory of the Aerospace Research Laboratories (1965-1975), the Department of Statistics of Iowa State University (1975-1995), and the Mathematical Sciences Department of the IBM T. J. Watson Research Center (1995-present). His research interests include mixed-effects linear statistical models, statistics in sports and competition, the design of experiments, and related topics. He has been elected a Fellow of the American Statistical Association (1975) and of the Institute of Mathematical Statistics (1980) and is an elected member of the International Statistical Institute. While at Iowa State, he served as the major professor (or co-major professor) for 14 recipients of the Ph.D. degree.

SCHEDULE

ABSTRACTS

The Selection or Seeding of College Football or Basketball Teams for Postseason Competition: a Statistician's Perspective
David Harville
Research Staff Member Emeritus
Mathematical Sciences Department
IBM Thomas J. Watson Research Center

Systems for ranking college football or basketball teams take many forms, ranging from polls to so-called computer ranking systems. Some of these are used in ways that have considerable impact on the teams. The committee responsible for the selection and seeding of teams for the postseason NCAA Division I men's basketball tournament is influenced by various rankings, including ones based on the RPI (Ratings Percentage Index). The BCS (Bowl Championship Series) rankings of NCAA Division I-A football teams determine which two teams compete in a postseason national championship game and determine eligibility for other prestigious postseason games. There are certain attributes that seem desirable in any ranking system that is to be used in selecting or seeding teams for postseason competition or that may have other tangible or intangible effects on the teams. The polls, the RPI, and the BCS rankings are notably deficient in some of these attributes. A system that incorporates all of the desired attributes (at least to a considerable extent) can be achieved by adopting a modified least squares approach in which the expected difference in score in each game is modeled as a difference in team effects plus or minus a home field/court advantage.

Varying Coefficient Tensor Models for Brain Imaging
Brian D. Marx
Department of Experimental Statistics
Louisiana State University

We revisit a multidimensional varying-coefficient model (VCM), by allowing regressor coefficients to vary smoothly in more than one dimension, thereby extending the VCM of Hastie and Tibshirani. The motivating example is 3-dimensional, involving a special type of nuclear magnetic resonance measurement technique that is being used to estimate the diffusion tensor at each point in the human brain. We aim to improve the current state of the art, which is to apply a multiple regression model for each voxel separately using information from six or more volume images. We present a model, based on P-spline tensor products, to introduce spatial smoothness of the estimated diffusion tensor. Since the regression design matrix is space-invariant, a 4-dimensional tensor product model results, allowing more efficient. computation with penalized array regression.

On Unbalanced Multistage Methodologies for the Partition Problem
Yuefeng Wu
Department of Mathematics
University of New Orleans

We consider the problem of partitioning a set of normal populations with respect to a control population into two disjoint subsets according to their unknown means. Taking c greater than or equal to 1 observations from the control population instead of the usual vector-at-a-time approach, we construct a two-stage and a purely sequential procedure. The theoretical second-order asymptotics of the purely sequential procedure are obtained. The performance of the two proposed procedures is studied via Monte Carlo simulations for small and moderately large sample sizes.

Inverse Prediction with Multivariate Response and Adjustment for Covariates
Lynn Roy LaMotte
LSU Health Sciences Center School of Public Health

A carrion-fly larva collected from a corpse may be used to estimate a minimum time since death. It has multivariate size measurements y* (for example, body length, dry weight, and length of the dorsal cornu), and it grew under conditions x* (e.g., degree-hours). Its age d* is unknown. Training data are available for y from multiple specimens of the same species over a grid of values of conditions x and ages d. Using interpolation models for the mean vector and variance-covariance matrix of y as functions of age d and the covariates x, we describe and illustrate a method for constructing an interval estimate of the age d of the mystery specimen, Two other settings are described, too. The method is based on a heteroscedastic multivariate mixed model for y, which extends our earlier work in Journal of Forensic Sciences, JFSCA, Vol, 40, No. 4, July 1995, 585-590,

A cDNA microarray analysis to detect differential gene expression between root types in sweet potatoes
Lisa Morris
Louisiana State University

Microarray analysis allows researchers to collect vast amounts of genetic data in very short time periods. A common application, the cDNA microarray, takes advantage of the uniqueness of mRNA sequences for individual genes and the cellular control mechanisms of gene expression to explore differential expression across different cell types, life or developmental stages, disease states, environmental stresses, or any of a variety of other treatment conditions. Until relatively recently, much of the research done using microarray analysis focused on extracting information about gene expression through clustering and other multivariate techniques. Newer methods are being developed using linear and mixed models to assess the differential expression of genes between cells at two or more levels of a given treatment effect. The enormous amount of data collected in a single microarray experiment, along with the many sources of variation inherent to the method, make the analysis of this type of experiment far more challenging than that for the other methods mentioned above. The primary purpose of this study was to identify which genes are expressed differentially between fibrous roots and storage roots in sweet potatoes and to state the nature of those differences (i.e. increased or decreased expression). Additionally, the researchers wanted to compare two different software packages used for data imaging and two different numbers of biological replications in the design of the experiment. This was the first microarray experiment performed in this laboratory, and the researchers hoped to develop a general protocol for future experiments. Loess regression was used for data normalization, and mixed model analyses were used to test for differential expression between the two root types. Additionally, the results of the mixed model analyses from the four different experimental designs were used as the dependent variable in a logistic multiple regression to test for differences between the software packages and number of biological replications used.