 STATISTICS SYMPOSIUM PROGRAM AND ABSTRACTS includes Mathematics Contributed Session
Welcoming Remarks: Friday, April 2, 2004, 8:45 am, CSM 217 Dr. Hector Flores, Dean of College of Sciences and Mathematics
Statistics Symposium Invited Session I: Friday a.m., CSM 217 Session Chair: Dr. Debra Ingram, Arkansas State University
9:00
10:00 a.m.
STATISTICAL RESEARCH AND GRADUATE EDUCATION IN THE 21ST CENTURY: INTERDISCIPLINARY RESEARCH Dr.
Jack D. Tubbs, Professor of Mathematics and Statistics, Director of the Institute
of This
talk will focus on the ever expanding role that statistics and statisticians
play at non Tier I research universities. As
smaller universities seek to improve their ranking among research institutions,
resources demand that these institutions improve their emphasis upon inter/intra
disciplinary research opportunities. This
emphasis not only affects the ways in which we evaluate our faculty but also the
ways that we train our students. I
will present the preliminary results of one such project which is motivated by
the growing widespread problem of water pollution with pharmaceuticals.
The work involves an analytical continuation of recent published studies
that indicated that the pharmaceutical fluoxetine (Prozac), a selective
serotonin reuptake inhibitor, is discharged in municipal wastewater treatment
plant effluents to surface waters. The
study attempts to answer some of the questions that are not presently
understood. They are:
(1) the magnitude, duration and frequency of fluoxetine exposure in
aquatic systems, (2) mechanistic toxicity of fluoxetine in non-target biota,
including behavioral responses, and (3) an assessment of environmentally
relevant fluoxetine concentrations is needed to characterize ecological
community responses.
Break 10:00 - 10:15 a.m.
PROCESS IMPROVEMENT USING THE
DMAIC METHOD Dr. Mike Jackson, Continuous Improvement Consultant,
Wal-Mart Stores, Inc. The
DMAIC methodology is a disciplined, data driven approach that focuses on a
process and, more specifically, the variation within a process. Many
companies have used this methodology to improve their business by eliminating
defects and reducing process variation. Although the methodology has been
used successfully in manufacturing for years, service organizations have
recently started implementing the philosophy. The five phases of this
improvement strategy are Define, Measure, Analyze, Improve, and Control.
Members of the Wal-Mart Continuous Improvement team are responsible for the
DMAIC training, project consulting, and statistical analysis. MODELING
THE INCIDENCE RATES OF POSITIVE DRUG TESTS IN THE AIR FORCE 1996-1999 Dr.
Chuck McGhee, Biostatistician, Tarrant County Public Health Department, The
Air Force noticed an increase in positive drug tests in 1999. They had
previously changed the criteria for random drug testing. This study looked
at the random drug tests from October 1996 to September 1999. The goal was
to predict positive drug tests and determine the trend of positive drug tests.
Logistic regression was used to determine risk factors for positive drug tests.
Cumulative incidence rates were tested for linear trend. Several variables
were significant (status, age, rank, transfer & turn around time, and
ethnicity) in the multivariate model. The cumulative incidence rate for
the 36-month time period increased by 0.05 and was significant. The
conclusions drawn from this were to continue monitoring the transport of samples
for testing and to target the younger enlisted personnel in all service groups
(active, reserve, and national guard) with more interventions. Statistics Symposium Invited Session II: Friday p.m., CSM 217 Session Chair: Dr. J. Edward Bennett, Emeritus Professor of Chemistry, Arkansas State University
MODELING, INFERENCE AND FORECASTING IN THE AGE OF FAST COMPUTERS Dr.
Katherine B. Ensor, Professor and Chair of Statistics, Advances
in statistical methodologies are inherently coupled to advances in computing
technology. Examples of such paradigm shifts in statistical modeling and
inference include the bootstrap, nonparametric function estimation, as well as
the age of the MCMC solutions to Bayesian models. Just as significantly,
the ability to easily simulate complicated stochastic processes provides us with
a means of utilizing stochastic models increasingly in our scientific
investigations. In this talk I will focus on the development of
methodologies for dependent processes including computer intensive advances in
inference and forecasting as well as highlight new work in the area of spatio-temporal
modeling.
2:45 3:15 p.m. FUNCTIONAL
COEFFICIENT AUTOREGRESSIVE MODELING OF MULTIVARIATE TEMPORAL DATA Dr.
Jane Harvill, Assistant Professor of Statistics, Univariate
nonlinear time series models have been used extensively over the past 15 to 20
years to model complex dynamic systems that cannot be adequately represented
using linear models. A very general type of nonlinear time series model is
the univariate functional coefficient autoregressive (FCAR) model, where the
autoregressive coefficients are allowed to change as a function of one or more
variables, which may be lagged values of the series or exogeneous predictors,
including, for example, time. We extend the univariate FCAR model to the
vector time series framework. A bootstrap test for vector time series
nonlinearity is presented. FCAR methods are applied to a number of
different datasets to illustrate utility. Extensions to spatial-temporal
modeling are mentioned, but remain an open problem. SOME SELECTION PROCEDURES FOR THE BIRNBAUM-SAUNDERS DISTRIBUTION Dr. Jeanne Hill, Visiting Assistant Professor of Statistics, In 1969, Z. W. Birnbaum and S. C. Saunders introduced a new fatigue life distribution which bears their names. This distribution was derived from considerations of the physical behavior of material subjected to cyclically repeated stress patterns. The distribution models the number of cycles needed to force the length of the fatigue crack to grow past a critical length. This talk will review some general results about the distribution and introduce new procedures for comparing two or more Birnbaum-Saunders populations. For two ore more independent Birnbaum-Saunders populations with unknown scale parameters, a better population is defined to be the one having a larger scale parameter. The problem of selecting t best of such populations is considered.
BAYESIAN STATISTICS:
AN IDEA WHOSE TIME HAS COME Dr. Brian J. Smith, Assistant Professor of Biostatistics, University of
Iowa
Today,
the Bayesian approach to statistical modeling enjoys widespread acceptance in
the statistical community. Not only an academic endeavor, the approach is
being applied to solve many real-world problems. Bayesian methods are
encouraged by the FDA for use in the development of medical devices and are
being used to build more effective spam filters for e-mail. The rise in
popularity of these methods is surprising considering that their use was one of
the most debated and controversial statistical topics no more than a decade ago.
We will look back at several advances that were key to the rise in popularity of
Bayesian modeling. Our talk will begin with the origin of Bayes Theorem
and then move on to the development of the Metropolis-Hastings algorithm, the
impact of high-speed computing, and the arrival of custom software for
performing Bayesian analyses. Statistics Symposium Invited Session III: Saturday a.m., CSM 217 Session Chair: Dr. Roger Abernathy, Arkansas State University
CONSTRUCTION OF OPTIMAL SCREENING DESIGNS FROM HADAMARD MATRICES Dr.
Debra Ingram, Assistant Professor of Statistics, Arkansas
Fractional
factorial designs are used in a wide variety of industrial and scientific
investigations, where time and resources are always limited, and, as a result,
run size economy and run size flexibility are important issues. Regular
fractional factorial designs are among the most widely used statistical
experimental strategies for studying the effects of several variables
simultaneously and are readily available in the literature. However, they
are limited in that the number of experimental runs must be a power of two,
leaving large gaps in the available run sizes. Nonregular
fractional factorial designs taken from Hadamard matrices can be constructed for
every run size that is a multiple of four, providing a distinct advantage over
the regular designs. This talk focuses on the use of computationally
efficient search algorithms applied to Hadamard matrices for the construction of
optimal nonregular designs. The new designs compare favorably with the
competing regular designs and make it possible for engineers and scientists to
plan experiments for many different combinations of run size and number of
variables to be studied. THE
LOG F: A DISTRIBUTION FOR ALL
SEASONS Dr.
Marty Spears, Associate Professor of Mathematics, A
family of statistical models based on the logarithm of an F variate was
introduced over 20 years ago. This family of distributions is extremely
versatile, but is little appreciated by the statistical community. The log
F family of distributions belongs in the tool box of applied statisticians and
should be used in data exploration. Examples of the log F model will be
presented that cover a variety of statistical functions and several application
areas. Details will be provided for where to obtain free computer code to
fit the models to data.
10:50 11:50 a.m. STATISTICS AS AN ENABLING
DISCIPLINE Dr. William B. Smith, Executive Director of the American
Statistical Association (ASA), Fellow of the American Statistical Association,
Professor Emeritus, Department of Statistics, Texas A&M University Real-world
examples and experiences will be presented in which effective use of statistical
approaches provided insight. These situations include the Firestone/Ford
tire failure analysis, an oil well pipe corrosion challenge, the recent
controversy over FBI chemical analysis of ammunition identification, a long term
investigation of a bone strengthening dietary supplement for race horses, and
advances in veterinary ophthalmology. Open research opportunities and
ASA's involvement in advancing the discipline will be shared. Mathematics Contributed Session: Friday p.m., CSM 217 Session Chair: Dr. J. Edward Bennett, Emeritus Professor of Chemistry, Arkansas State University KNOTS, BRAIDS, AND DNA Dr. Grant Lathrom Department of Mathematics and Statistics, Arkansas State University, PO Box 70, State University, AR 72467 Knot
theory has applications in many areas of physical science as well as being a
fascinating area of mathematical research on its own.
We will give a brief overview of knot theory and describe some techniques
for determination of polynomial invariants for knots.
We will conclude with a discussion of the application of knot theory to
DNA replication. DISTINCT PATTERNS OF HEMI-SPATIAL NEGLECT: A MULTI-LESIONED MATHEMATICAL
MODEL Tony Sloan, M.D. Department
of Family Medicine, University
of Left
hemi-spatial neglect typically occurs after a right hemispheric stroke. Patients
with hemi-spatial neglect exhibit a range of bizarre but consistent behaviors on
classic clinical tests. Patients to some degree fail to attend to left
hemi-space or to left sides of objects. A multi-lesioned mathematical model, the
Gestalt grouping-combined hemi-attentional-affine representational model, is
used to show the existence of several distinct patterns of hemi-neglect behavior
that exist empirically in the neuroscience literature. Moreover, we establish
the existence of several more distinct patterns of hemi-neglect behavior that
have yet to appear in the empirical literature. The space of external visuo-spatial
stimuli is realized as the space, H([a,b] Ũ P), of compact subsets of a strip,
[a,b] Ũ P, in the plane Pē. A Gestalt grouping procedure decomposes a finite
collection of stimuli into disjoint perceptual objects. An internal
representation of visuo-spatial stimuli is determined by the action of a group
of isometries on H([a,b] Ũ P). A combined hemi-attentional system, given by a
linear combination of a pair of normal distributions, is applied to the internal
representation. A lesion can be a representational error (related to the
isometries), an attentional error (related to the combined hemi-attentional
system), or a Gestalt grouping error. |