9:00 Introduction

9:10 QUANTIFYING THE PREDICTIVE INFORMATION OF HEALTH-RISK BEHAVIORS

Todd G. Nick*, LaKeshia Lawrence, and Jenelia Bracey, University of Mississippi Medical Center, Jackson, MS 39216, and Murrah High School, Jackson, MS 39216

The Youth Risk Behavior Surveillance System (YRBSS) monitors six sets of priority health-risk behaviors among adolescents. These categories include behaviors that contribute to injuries, tobacco use, alcohol and other drug use, sexual behaviors, unhealthy dietary behaviors, and physical inactivity. Quite frequently, many candidate predictors within one of the categories are regressed on a response from another category and stepwise selection procedures are utilized. If the candidate predictors are from different categories, then researchers mistakenly perform two or more multiple regression analyses on the different categories. Also, little attention is given to assessing the contribution of the entire set with multiple degree of freedom tests. This talk focuses on applying multiple degree of freedom tests and incorporating measures of adequacy [Harrell (1985)] to compare competing categories that each describe a response. Logistic regression models are used on the YRBSS to illustrate these methods. This block analytic approach provides an a priori analysis and ensures the necessary logic for appropriate data analysis. It compares favorably against stepwise selection procedures and along with a measure of adequacy of information offers easily interpretable models.

9:30 A CORROBORATION OF HAND AND COMPUTER CALCULATIONS FOR A SELECTED SAMPLE OF INTRODUCTORY STATISTICS HOMEWORK ITEMS

Bryan K. White* and Richard F. Ittenbach, University of Mississippi, University, MS 38677

The professional educational literature has examined computer-assisted instruction and teacher-directed instruction, often in a compare and contrast context. It is possible that a multidirectional approach to learning using elements of both computer-assisted instruction and teacher-directed instruction has implications for the general student population. The purpose of this paper is to illustrate how computer-assisted instruction combined with teacher-directed instruction can work toward the benefit of students by increasing competency in both statistics and the use of computers.

9:50 MULTIVARIATE REGRESSION MODELS WITH NESTED ERROR STRUCTURE

Pradeep Singh* and Govinda J. Weerakkody, Mississippi State Unversity, MS 39762

We consider the likelihood ratio criterion for regressor parameters in multivariate regression models with nested error structure. Following Fujisawa (1997) and Anderson et al. (1986), we present bounds and an approximate distribution of the likelihood ratio criterion.

10:10 Break

10:20 MODELING THE BODY SURFACE AREA OF DOGS USING REGRESSION GRAPHICS

Carolyn R. Boyle, Mississippi State University, Mississippi State, MS 39762-9825

Body surface area (BSA) is important to physiologists and clinicians because it is strongly associated with numerous physiologic variables. Because it is difficult to measure BSA directly, many attempts have been made to relate BSA to other body measurements such as volume, weight, or height. Previous work by others considered nonlinear models that were linearized by applying a logarithmic transformation. An alternative approach is to develop a linear model directly using the modeling process proposed by Cook and Weisberg in their 1994 text An Introduction to Regression Graphics. Their method uses graphical displays to explore the conditional relationship between the vector of predictors and the response variable. The goal is to determine the smallest number of linear combinations of the predictors that contain the same information about the response as the original predictors did. If necessary, the original predictors are transformed so that the relationships among the predictors are all linear. Cook and Weisberg show how the marginal distributions of these "linear predictors" reveal the structural dimension of a regression problem and permit estimation of the ideal summary plot. My research used regression graphics to model the relationship between BSA and various measurements of body length and circumference using data from 42 dogs. The adequacy of these models and that of traditional models was assessed using graphical techniques recently published by Cook and Weisberg (JASA 92(438): 490-499, 1997).

10:40 SMOOTHED LAD ESTIMATION IN LINEAR REGRESSION MODELS

Mi-Young Jang* and Zhijun Liu, Mississippi State University, Mississippi State, MS 39762

The least absolute deviations (LAD) method as a robust alternative to the least squares (LS) method in statistical inference has attracted considerable research interests. Research on asymptotic properties and computation techniques of the LAD in linear regression models is fruitful. However, computation is still a problem for a model with large number of independent variables. The smoothed LAD method is introduced in this work to tackle this problem. Ideas and asymptotic properties of the smoothed LAD will be presented.

11:00 A MULTIMEDIA TOUR OF AN ARCHAEOLOGICAL SITE IN EK BALAM, MEXICO

Brian P. Hughes* and Andrew V. Royappa*, Millsaps College, Jackson, MS 39210

Multimedia incorporates pictures, text, audio, and video and is used increasingly in education. Using several multimedia components, we have created a program that allows the user to view an archaeological excavation in Ek Balam, Mexico. The program consists of a multimedia "walk-through" with static images and two virtual-reality type environments: one based on 360 degree panoramas reconstructed from photographs, and another based on 3D scenes created using modeling tools and expressed in VRML. The entire program can be delivered over the WWW and utilizes several multimedia & graphics technologies, including Authorware, Shockwave, QuicktimeVR, TrueSpace, and CorelCAD. The program is useful for helping students and others to learn more about Ek Balam. It is intended for disseminating information about the archaeological excavation using cutting-edge multimedia tools, and we hope that it will be useful for teaching classes in archaeology, anthropology, and sociology at both the undergraduate and K-12 levels.

11:20 Divisional Business Meeting
 
 

1:30 NUMERICAL SOLUTION OF PDES USING REGULARIZED APPROXIMATIONS

Joseph Kolibal, University of Southern Mississippi, Hattiesburg, MS 39406

The numerical solution of partial differential equations in which spatial derivatives are approximated using mollifyers of the discrete data is developed. The technique is applied to the solution a model hyperbolic pde based on the delaval nozzle problem for inviscid flow. The approximation yeilds a quasi-analytical approximation to the derivative which has advantages for adaptive methods.

1:50 THE ROLE OF DISEASE INCUBATION PERIOD STUDIED IN A DETERMINISTIC MATHEMATICAL MODEL

John R. Graef, Michael Y. Li*, and Liancheng Wang, Mississippi State University, Mississippi State, MS 39762

The possible roles played by the incubation period in the disease transmission within a population are studied in a differential equation model. The basic assumptions are that the disease spreads through direct contact among hosts and it may cause death. Our models show that, in a slowly growing host population of large size, a disease with a long incubation period tends to become endemic whereas one with no or short incubation period will die out from the host population. Certain pathogens (e.g., HIV virus and TB bacillia) are know to have variable incubation periods in human hosts ranging from a few days to life time. Our results seem to imply that, in a large and slowly growing population (e.g., in China), mutations of such a pathogen that possess a long incubation period would prevail and become dominant in the population. This can be useful for public health workers in designing appropriate disease-control measures.

2:10 BOUNDARY CONDITION EFFECTS ON 2-DIMENSIONAL WAVE INTERFERENCE

Alice Qiao¹*, Mike Zhou¹, and A. Louise Perkins², ¹University of Southern Mississippi, Hattiesburg, MS 39406, and ²Naval Research Laboratory, Stennis Space Center, MS 39529

In this work we do numerical simulations of wave interference on a 2-dimensional square domain. Various boundary conditions are tested. It is shown that the interference pattern is quite sensitive to boundary conditions. Due to reflection, Dirichlet boundary conditions lead to large errors in the inner domain. The direct use of a theoretical approximation along the boundary also gives poor results. In contrast, a modified radiation boundary condition is shown to work well for a nested internal boundary.

2:10 Break

2:45 ON A MIXED IMPLICIT/EXPLICIT TIME INTEGRATION ALGORITHM

Hench Qian* and Jianping Zhu, Mississippi State University, Mississippi State, MS 39762

Time dependent partial differential equations are solved using either explicit or implicit integration algorithms. Explicit time integration algorithms are easy to implement, particularly for high dimensional and nonlinear problems, and well suited for parallel computers. However, they are only conditionally stable in most cases. On the other hand, implicit algorithms are unconditionally stable, but more difficult to implement for high dimensional and nonlinear problems, and harder to parallelize on multiprocessor computers. In this paper, we will discuss a mixed implicit/explicit algorithm for multiprocessor parallel computers. The basic idea is to decompose the original problem domain into subdomains and integrate each subproblem using implicit algorithms. At the boundaries of subdomains, the numerical boundary conditions are treated explicitly, so that subproblems in different subdomains are independent of each other and can be solved in parallel. The solution values at the boundaries between subdomains are updated using a predictor-corrector (explicit/implicit) type algorithm to maintain unconditional stability of the time integration algorithm. Detailed stability analysis and numerical examples will be given in the presentation.

3:00 TOTAL SIMULATION ERROR

Peter A. Orlin* and A. Louise Perkins, Planning Systems, Inc., Slidell, LA 70458, and Naval Research Laboratory, Stennis Space Center, MS 39529

Finite difference methods for approximating derivatives were constructed using weighted averages of different stencil widths. The weights (undetermined coefficients), determined in frequency space, satisfy a priori phase accuracy requirements. This allows us to reduce actual (not just formal) numerical truncation error, especially for bandlimited processes. We found that, even with phase accurate approximations, long-time integration simulation error may still grow large. This is because the total simulation error (TSE) also includes the physically-based mathematical approximation error in the governing equations.

3:20 SIGN K-POTENT SIGN PATTERN MATRICES

Jeffrey Stuart*, Carolyn Eschenbach, and Steve Kirkland, University of Southern Mississippi, Hattiesburg, MS 39406, Georgia State University, Atlanta, GA 30303, and University of Regina, Regina, Saskatchewan, S4S 0A2

The sign pattern matrix is called sign k-potent if k is the smallest positive integer for which the ^k+1 power of equals . We characterize the irreducible pattern matrices that are sign k-potent. For a reducible pattern matrix in Frobenius normal form, we determine necessary conditions for when A is sign k-potent in terms of the structure of the irreducible, diagonal blocks.

9:00 USING ESTIMATES OF MEANS AND VARIANCES TO ASK QUESTIONS

Deborah L. Boykin, USDA Agricultual Research Service, Stoneville, MS 38776

Part of the initial research process is to restate the research objectives as questions or hypothesis. Answers to these questions should allow the researcher to reach objectives of the study. This presentation will describe how estimates of means and variances can be used to form hypothesis and seek answers to questions. T-test and F-test are basic tools for addressing questions. To use these tools, one must obtaining estimates of means that are free from bias and obtain the best estimate of error. Analysis of variance can be used to construct F-test and a corresponding estimate of error. Also, mean comparison procedures are often used to help the researcher answer questions. Several popular mean comparison methods will be discussed.

9:20 BOOTSTRAP CONFIDENCE INTERVALS IN LOCATION MU-ESTIMATION

Zhijun Liu* and Mi-Young Jang, Mississippi State University, Mississippi State, MS 38677

The Mu-estimation is created by incorporating ideas of the M-estimation and U-statistics. Asymptotic properties have already explored by Liu (1997). Bootstrap method is used for producing good approximation of confidence intervals. This paper is focused on bootstrap confidence intervals in location Mu-estimation.

9:40 AN INTEGRATED CONCEPTUAL MODEL OF RESEARCH DESIGNS AND METHODS USED IN THE SOCIAL AND BIOBEHAVIORAL SCIENCES

Richard F. Ittenbach, University of Mississippi, University, MS 38677

Although discussions of research methods and designs tend to be dominated by quantitative perspectives, recently, a number of qualitative methods have established their place in the published literature. From observational to experimental and from single-subject to large-group, it may well be that a design that incorporates both quantitative and qualitative components offers a broader, more balanced approach to hypothesis testing and problem resolution. The purpose of the present paper will be to present a concise, integrated conceptual model that unites under one framework both quantitative and qualitative designs and methods.

10:00 Break

10:10 RANDOM LITTER SIZE PROCEDURES FOR DEVELOPMENTAL TOXICITY STUDIES

Dale Bowman, University of Mississippi, University, MS 38677

In a typical developmental study, primary interest is often on estimating risk associated with exposure to a toxic substance. Exposure may effect death rates, malformation rates and birth weight of fetuses. Exposure may also effect the size of the litters, although most procedures which are commonly used to estimate risk condition on litter size. Procedures are developed here which treat litter size as a random variable. The effects of conditioning on litter size are explored and procedures are developed for estimating risk associated with binary endpoints often measured in the presence of random litter sizes modeled as functions of dose.

10:30 EFFECTIVENESS OF ACT SUBSCORE IN MATHEMATICS AS A PREDICTOR OF STUDENT SUCCESS IN CALCULUS I

Govinda J. Weerakkody, Mississippi State University, Mississippi State, MS 39762

Linear relationship between student success in Calculus I and ACT subscore in mathematics controlling for the possible effects of gender, race, and section is investigated. Two measures of success, the departmental final exam score and the earned letter grade are considered. Mixed linear models approach using SAS Mixed procedure is used in the analysis of the departmental final exam score. A Generalized mixed linear models approach using the GLIMIX SAS macro is used in the analysis of the earned letter grade.

10:50 Business Meeting of the Mississippi Chapter of the American Statistical Association

1:30 THE COMPONENTS OF A GOOD WEB PAGE

E. Scotty Scott II* and Gwang Jung, Jackson State University, Jackson, MS 39217

My research includes finding out what exactly makes a web page good or what makes one bad. I have found that creating a good page is similar to writing a good essay. The most important step in creating a good web site is determining a topic. Once a topic has been found, you must remain on that subject. If you remain on topic, then you have a good web site. However, a lot of people choose to put huge amounts of garbage on their page that has absolutely nothing to do with the site's original purpose. The next phase is putting together a color scheme. The colors that are chosen must not clash with each other. Instead, they must complement each other. If that happens, then the page is not pleasing to the eye. Now, the next phase my research instills is the use of CGI Scripts. This technique has many uses. Some of the most common ones are: processing forms, image maps, animations, counters, search engines, and WAIS servers.

1:50 INCREASING WEB INTERACTIVITY WITH DYNAMIC HTML

Vijay Amirineni* and Jerry Kandies*, Delta State University, Cleveland, MS 38733

Common Gateway Interface (CGI) scripts and Java applets provided the web developers with the much needed capability to create interactive features. This interaction was lacking in the web pages coded with the previous versions of HTML. One of the limitations of using CGI scripts is that they require extra processing time from the server and thus add an overhead on the server. Java programming, on the other hand, confronts the part-time programmers with a steep learning curve. Dynamic HTML is going to change all that and dazzle the interest and curiosity of the web surfers with its interactive capabilities. Dynamic HTML will make it possible to create web pages with images that a user could drag from one location to the other. The user could visit a virtual furniture store on the web, arrange the furniture in a virtual room according to his/her tastes and get a bird's eye view before actually paying for it. Dynamic HTML allows the programmer to create menus so as to better organize the contents in a web page. The added features that dynamic HTML brings are endless and these extended features are sure to generate a whole new dimension of web programming and how the web pages are viewed. This paper talks in depth about the important features and differences between the Netscape Communicator 4.02 and Microsoft Internet Explorer 4.0 browsers, and how they provide life or interactivity to the web pages using the new Javascript 1.2.

2:20 EVERYONE LOVES THE BOOM

Gerard L. Howard, Jackson State University, Jackson, MS 39217

My project was to design a web-site for the universities marching band ("The Sonic Boom of the South"). I set up a questionnaire page for visitors asking: what type of surfers they were: students or otherwise, alumni/current band members, and how did they found out about the site. My hypothesis was that more people would visit during football season since that was the time the band was most active. I also hypothesized that most of the visitors to the site would be members of other university marching bands. Unfortunately due to a miscommunication with one of the schools system administrators most of my information collected during the summer was lost. Since then I've collected 13 responses from visitors to the site. My hypothesis in both cases was correct. Further results will be discussed during the presentation of the project.

2:40 YEAR 2000 SOFTWARE GLITCHES AND THE PROGRAMMER ADVANTAGES

Vijay Amirineni, Delta State University, Cleveland, MS 38733

The Information Systems community is headed toward a devastating crash in the year 2000 because of our faulty standard date format: MM/DD/YY. This faulty programming standard, otherwise known as the 'millennium bug,' is costing the worldwide computer community billions and billions of dollars as the programmers are charging anywhere between $1.00 - $2.50 per line of code that they browse to fix it. The Fortune 500 companies are expected to spend $50 million to $100 million each to fix their information systems. The problems started when programmers decided to use only the last two digits of the year, such as 69 instead of 1969, to save storage space while recording the dates. If we enter 00 for the year 2000, the programs may view it as year 1900, thus causing widespread problems like system crashing, corrupted records, unsatisfied customers, and lost businesses. Like adding insult to injury, law suits from customers may aggravate the agony. Detroit is already credited with the first Year 2000 related lawsuit. This paper talks in depth about the issues relating to the millennium bug; how can the companies face the problem and what is already being done by some major corporations; and how can we, the computer programmers, prosper from this commotion.