Fall 2000

[12/14 | 12/1 | 11/20 | 11/3 | 10/26 | 10/16 | 9/25 | 9/11 ]

12/14/00 2:00 - Your Turn, My Turn, Whose Turn? Packet Scheduling in Wireless Networks
Wesley Murry
Dickinson College

Wireless data networks are becoming an integral part of the Internet, expecially as a network access technology (i.e. PDA's, Laptops, etc). In response, research has centered on extending the quality-of-service (QoS) and fairness models for wired networks to wireless networks. My research has centered on learning about packet scheduling algorithms in the wireless network model. It is the hope that through analysis of these algorithms a metric of fairness can be developed.

12/1/00 - Anxiety Over Statistics (or Vice Versa?)
Bob Starbuck
Wyeth-Ayerst

A clinical trial of an anxiolytic agent will be presented to illustrate some of the topics and issues that confront a statistician involved in clinical research in the pharmaceutical industry. Topics that will be covered include clinical trial protocol design, sample size estimation, randomization, handling of missing values, grouping of data, data analysis, and presentation of results. Career opportunities and success factors will also be discussed.

11/20/00 - Comparing and Evaluating Bayesian Belief Network Learning Algorithms When Applied to Predicting Stock Price
Carol Wellington
Department of Computer Science
Shippensburg University

A Bayesian belief network (BBN) is an artificial intelligence model used in applications that require reasoning under uncertainty. While the models have been hand-built successfully for a number of years, current research is attempting to "learn" the model from historic data instead of relying on human experts for model construction. After detailing the issues these learning algorithms are designed to handle, I will show results of applying my learning strategy to the prediction of stock prices. Now that I have models resulting from various learning algorithms, I am working on how to evaluate learning algorithms in this application. I will show the way learning algorithms have traditionally been evaluated, why that is not particularly insightful in the application and some other alternatives I have been investigating. In some ways, the learned model could be considered to be a new technical analysis tool, so the evaluation methods I am investigating could be applied to evaluate and compare the usefulness of other technical analysis tools.

Fri: 11/3/00 12:00 - Creatively Desperate (or Desperately Creative) Statistics: A Case Study of Repeated Measures Analysis with Sparse Data
Thomas H. Short
Department of Mathematical Sciences
Villanova University

As an applied statistician, I have the opportunity to work on many interesting and challenging analyses. One of my favorites involved tracking cognitive development in Villanova Nursing students over the their four years of college. An instrument known as the Learning Contexts Questionnaire (LCQ) was supposed to be administered to each student in each semester, but many of the values in the dataset are missing. Without the missing values, I would have performed a standard repeated measures analysis. The missing values forced me to think creatively to develop an analysis that would be true to the data and also understandable to my client.

Thurs: 10/26/00 12:30 - `God help the state of Maine when mathematics reach for her': integer allocation and the Alabama paradox.
Jim Wiseman
Northwestern University

The U.S. Constitution provides that seats in the House of Representatives "shall be apportioned among the several states according to their respective numbers," but doesn't give any details on exactly how this is to be done. The history of apportionment includes lots of political backbiting and surprisingly rich mathematics. I'll answer such burning questions as "Why are there 435 members of the House?" and "How *are* they apportioned among the states?"

10/16/00 - Tea for Two and T2-Spaces: The Point of Point-Set Topology
Thomas L. Drucker

As we enter the 21st century, it's worth looking back at a confluence of factors that created a new discipline around 1900. The familiar geometry of Euclid was being generalized to include geometries that stretched familiar space in different directions. The points that had been problematic for mediaeval philosophers talking about angels were now made equivalent to functions. Finally, Georg Cantor was demonstrating that set theory could be robbed of some of its mystery and domesticated for mathematical use. Out of this mixture was born the discipline we call set-theoretic topology. No background in set theory or topology is assumed but we shall say a little about why the Erlangen program lasted longer than most programs in the new television season are likely to.

9/25/00 - Being an Intern in .com World
William Menzie
Dickinson College

William, a physics major at Dickinson, spent his summer doing an internship with the CA Internet startup SearchButton.com. He will tell us about his internship, how he found it and the types of things that he did and learned.

9/11/00 - Bees, Ants, Robots and Wireless Networks
Grant Braught
Dickinson College

How are bees, ants, robots and wireless networks related? We'll explore the relationship between social insects, colonies of robots and the use of wireless networks for communication. We'll see how real ants solve shortest path problems and how this can help with network routing. Finally we'll look at a routing algorithm based on bee pollination that I developed for routing data in an Ad-Hoc wireless network.