AMERICAN STATISTICAL ASSOCIATION
February 9 – 10, 2001
University of North Florida
Friday, February 9, 2001College of Health Building, Room 101612:05 pm Registration
SESSION I: 12:45 pm – 2:50 pm
(Student Competition Paper Session)
Chair William J. Wilson, University of North Florida
12:45 – 1:00 pm Opening Remarks
Henry Camp, Dean, College of Arts and Sciences
University of North Florida
1:00 – 1:35 pm “2001: Statistical Odysseys in the Information
Age.”Richard L. ScheafferPresident, American Statistical Association
1:35 – 2:00 pm “Detection of locations of lightning events
on the basis of ground-based measurements”Marianna Pensky and Aicha ElhorDepartment of MathematicsUniversity of Central Florida
2:00 – 2:25 pm “The use of empirical likelihood in survival
analysis”Ian W. McKeague and Yi-chuan ZhaoFlorida State University
2:25 – 2:50 pm “Random Matrices With Dirichlet Distributed
Elements.”P. Rumchevaa, Tz. IgnatovbDepartment of Statistics, University of FloridaDepartment of Statistics and Econometrics,
Faculty of Economics and Business
2:50 – 3:00 pm Session Break
SESSION II: 3:00 – 5:20 pm
(Research and Methodology Session)
Chair Pali Sen, University of North Florida
3:00 – 3:20 pm “Predicting Y from a discretized version of X.”Jayaram SethuramanFlorida State University
3:20 – 3:40 pm “24-Run Split-Plot Experiments For Robust
Parameter Design”Scott KowalskiUniversity of Central Florida
3:40 – 4:00 pm “Breast Cancer Detection Using Rank
Nearest Neighbor Classification Rules”Subhas BaguiUniversity of West Florida
4:00 – 4:20 pm “Option-3 Median Measurement Scheme
in Outlier Reduction”Mark C. K. YangUniversity of Florida
4:20 – 4:40 pm “Robust I-Sample Analysis of Means Type
Randomization Tests for Variances”Anthony J. BernardUniversity of North Florida
4:40 – 5:00 pm “Post-Sample Stratification in Residential
Survey Design with Application to Economic
Impact of 'Daytrippers' on Daytona Beach”Mark D. Soskin, UCF
5:00 – 5:20 “Improved Multiple Comparisons with the
Best in Response Surface Methodology”Laura MillerUniversity of North Florida
5:30 – 6:15 pm ReceptionCourtesy of John Wiley & Sons
6:30 pm Banquet (Foundation Board room in the Arena)
Special recognition of Mike Bundrick by FLASA
Speaker introduction: Peter Wludyka, University of
Guest Speaker: George Casella, University of Florida
Saturday, February 10, 2001College of Health Building, Room 1016
8:30 – 9:00 am Continental Breakfast
(Courtesy of Addison Wesley Longman)
SESSION III: 9:00 am – 11:05 am
Chair James Gleaton, University of North Florida
9:00 – 9:10 am Presentation of the Student CompetitionWinner AwardJayaram Sethuraman, Florida State University
9:10 – 9:55 am “A Capstone Course for Undergraduate
Statistics Majors”John D. SpurrierUniversity of South Carolina
9:55 – 10:05 am Short Break
10:05 – 11:05 am Panel Discussion: “Teaching Statistics in the21st Century”
Fred Leysieffer, FSU Duane Meeter, FSU Kenneth Portier, UF Moderator: Mark Soskin, UCF
11:05 am Business Meeting (Room 1001)Jayaram SethuramanPresident, Florida Chapter ASA
ABSTRACTS 2001: STATISTICAL ODYSSEYS IN THE INFORMATION AGE Richard L. Scheaffer President, American Statistical AssociationThe age of data and technology now upon us should be a marvelous time for statistics, but successfully questing the trails that lie ahead is going to require concerted energy and careful planning. ASA can help, but ASA is essentially an organization of volunteers. Thus, any success that ASA might have in enhancing statistics education and practice is, in large measure, due to the commitment and expertise of the many members who donate much time and effort to its programs. This will continue to be true in the future, as we work together on issues such as the effect of information technology on publications, continuing education of statisticians, outreach education to other professional fields, education of future K-12 teachers, meeting the needs of applied statisticians (perhaps through certification), broadening the interest in Sections, activating Chapters throughout the land, and building an efficient but effective committee structure. This presentation and discussion will focus on key strategic initiatives for the new century. The formulation and implementation of programs to meet new challenges, as well as the ultimate success of those programs, depends on input and interest from a broad spectrum of the membership. Together, we can make ASA an effective voice for statistics in the new century.
Detection of locations of lightning events on the basis of ground-based measurements
Marianna Pensky and Aicha Elhor
Department of Mathematics University of Central FloridaThe problem of detection of locations of lightning events on the basis of ground-based measurements has been studied extensively within the last three decades. The location of a lightning event is derived from the times of arrival of electromagnetic radiation at several locations. The differences in the times of arrivals are converted into differences in distances from the point of origin (x,y,z) of the radiation to (m+1) receiving sites located in xy-plane at (a_i, b_i, 0), i=0,...,m. The objective of the present paper is to apply Bayesian estimation with various noninformative priors: the Jeffreys's prior and the reference and the reverse reference priors which coincide in the case of the above problem. Using Monte Carlo simulations, we compare the performance of the Bayesian estimators based on Jeffreys's prior, reference prior and flat prior with the performance of the deterministic method based on solution of a system of linear equations.
The use of empirical likelihood in survival analysis
Ian W. McKeague and Yi-chuan Zhao
Florida State UniversityThe use of empirical likelihood in survival analysis goes back to Thomas and Grunkemeier (1975) who derived pointwise confidence intervals for the survival function. In the last ten years the method has been applied to a variety of statistical problems. In this talk we apply this approach to compare survival functions for k-sample problems in survival analysis. n the first part of this talk we derive a simultaneous confidence band for the ratio between survival functions based on independent right-censored data. Earlier authors have studied such bands for the difference of two survival functions, but the ratio provides a more appropriate comparison in some applications, e.g., in comparing two treatments in biomedical settings. Our approach is formulated in terms of empirical likelihood and avoids the use of simulation techniques that are often needed for Wald-type confidence bands. The proposed methods are illustrated with a real data example. In the second part of this talk we consider ratios of cumulative hazards functions, i.e. cumulative hazard ratios. Pointwise confidence bands/tubes are asymptotically distribution-free in all cases.The cumulative hazard ratio provides a more appropriate comparison among treatments than difference of cumulative hazard functions because it is scale invariant.
Random Matrices With Dirichlet Distributed Elements
P. Rumchevaa, Tz. Ignatovb
Department of Statistics, University of Florida Department of Statistics and Econometrics, Faculty of Economics and Business Administration, Sofia UniversityMotivated by the Ergodic Theorem for Markov Chains, we examine the stochastic behavior of the product of a random vector that has Dirichlet distribution, with a random matrix which rows are independent and also have Dirichlet distributions. The exact distribution of this product is derived under different conditions about the parameters of the Dirichlet distributions.
Predicting Y from a discretized version of X
Florida State UniversityThis problem arose from a question raised by scientists at the DOD laboratory. Let (X,Y) be a pair of random variables and suppose that we want to predict the dependent variable Y from the independent variable X. Break up the range of X into k intervals and define the discretized variable X' as equal to 1, 2, ... , k on those intervals. The question that we ask is this. Can X' predict Y as well as X can? What is the loss in predictive ability? For a fixed value of k, what is the best way to divide the range of X into k intervals? We describe the optimum way to divide the range of X; this optimum possesses some surprising properties for which we do not see an obvious intuitive explanation. We give a complete solution for the case when (X,Y) is bivariate normal. When we heard of this problem, we conjectured that the optimal solution would divide the range of X into k intervals of equal probability. From the general results above, and from the particular results for the bivariate normal,we see that this is not the correct answer.
24-RUN SPLIT-PLOT EXPERIMENTS FOR ROBUST PARAMETER DESIGN
University of Central Florida
Split-plot experiments where the whole plot treatments and the subplot treatments are made up of combinations of two-level factors are considered. Often times in industry, due to time and/or cost constraints, the size of the experiment needs to be kept small. Many of these experiments can be thought of as robust parameter designs involving noise factors and design factors. Using fractional factorials and confounding, sixteen run designs are constructed that make efficient use of the available resources. Along with this, semifolding is used to add eight more runs. The resulting 24 run design breaks some of the alias chains and provides some degrees of freedom for a subplot error variance. Also, an alternative 24-run design based on the balanced incomplete block design is presented.Key Words: Aliasing, Design of Experiments, Fractional Factorials, Split-Plot Experiments
BREAST CANCER DETECTION USING RANK NEAREST NEIGHBOR CLASSIFICATION RULES
Subhash C. Bagui
The University of West FloridaIn this article we propose a new generalization of the rank nearest neighbor (RNN) rule for multivariate data for diagnosis of breast cancer. We study the performance of this rule using two known databases and compare the results with the conventional k-NN rule. We observe that this rule performed remarkably, and the computational complexity of the proposed k-RNN rule is much less than the conventional k-NN rule.
Option-3 Median Measurement Scheme in Outlier Reduction
Mark C. K. Yang University of FloridaWhen measurements are subject to rare but large errors, it is better to measure twice instead of once, and if the two measurements differ too much, take a third measurement. This is called the option-3 scheme. This talk will discuss some recent development of this scheme, showing that using the median of the three measurements is better than using the average of the two closest. To reach the maximum sample size benefit, the threshold for taking the third measurement is approximately 3 times the measurement error standard deviation.
Robust I-Sample Analysis of Means Type Randomization Tests for Variances
Anthony J. Bernard University of North FloridaFour Analysis of Means (ANOM) type randomization tests for testing the equality of I variances are presented. Randomization techniques for testing statistical hypotheses can be used when parametric tests are inappropriate. Suppose that I independent samples have been collected. Randomization tests are based on shuffles or rearrangements of the (combined) sample. Putting each of the samples “in a bowl” forms the combined sample. Drawing samples “from the bowl” forms a shuffle. Shuffles can be made with replacement (bootstrap shuffling) or without replacement (permutation shuffling). The tests that are presented offer two advantages. They are robust to non-normality and they allow the user to graphically present the results via a decision chart similar to a Shewhart control chart. The decision chart facilitates easy assessment of both statistical and practical significance. A Monte Carlo study is used to identify robust randomization tests that exhibit excellent power when compared to other robust tests.
Post-Sample Stratification in Residential Survey Design with Application to Economic Impact of 'Daytrippers' on Daytona Beach
Mark D. Soskin
University of Central FloridaTourist-based areas such as Daytona Beach are often popular "daytripper" destinations by nearby population centers. Managing environmentally-sensitive recreational resources such as Daytona require detailed socioeconomic survey data of these destination sites. This paper describes a survey design methododology that overcomes the traditional barriers to unbiased data collection. This design is then applied to a recent mail survey of Central Florida household beach visitation preferences and annual expenditure patterns. Spending and preference estimates from this survey paint a startlingly compelling directive for regional beach policy.
“Improved Multiple Comparisons with the Best in Response Surface Methodology”
University of North FloridaIn Response Surface Methodology (RSM), the objective is to determine the optimal point of the predictor variables, within the operability region, where the response is optimized. But, it is possible this point is not a reasonable option due to practical considerations, such as expense. In this situation, multiple comparisons can be performed with other points in the region to determine if some other points provide responses that are not significantly different from the optimal point.Moore and Sa (1999) approached this problem by constructing simultaneous confidence intervals for for all and referred it as multiple comparisons with the best in Response Surface Methodology. This article is trying to address the same problem but to restrict the region of interest to within a sphere with radius . A small scale efficiency study shows that the approximate sample-size saving of the improved method over the Scheffé method ranged from 34% to 47%.
A Capstone Course for Undergraduate Statistics Majors
John D. Spurrier
Professor of Statistics, Department of Statistics, University of South CarolinaThis presentation discusses a capstone course for undergraduate statistics majors at the University of South Carolina. The course synthesizes lessons learned throughout the curriculum and develops students' nonstatistical skills to the level expected of professional statisticians. Student teams participate in a series of inexpensive laboratory experiments that emphasize ideas and techniques of applied and mathematical statistics, mathematics, and computing. They also study modules on important nonstatistical skills. Students prepare written and oral reports. If a report is not of professional quality, the student receives feedback and repeats the report. All students leave the course with a better understanding of how the pieces of their education fit together and with a firm understanding of the communication skills required of a professional statistician.
Panel Discussion: “Teaching Statistics in the 21st Century”
Panelists: Fred Leysieffer, FSU Duane Meeter, FSU Kenneth Portier, UF Moderator: Mark Soskin, UCF The panel discussion questions will include (1) statistics education in a hi-tech classroom, (2) making statistics relevant to 21st Century problems, and (3) challenges for a web-supported statistics class.