Business and Economics Data Analysis Competition for Undergraduates

Each of the last four years, undergraduates in all disciplines from all colleges and universities in the United Sates and Canada have been invited by the American Statistical Association to participate in an Undergraduate Data Analysis Contest. Winners are recognized at the Joint Statistical Association annual meetings and cash prizes are awarded. This annual contest has proven to be popular among students of statistics, and faculty recognize that the contest produces a richer experience than that provided by textbook problems or case studies.

Unfortunately, the data set chosen for the contest is not always appropriate for students from all disciplines. For instance, the data set for the 1998-1999 contest involves predicting whether a mushroom is poisonous or not. Predictions are to be made based on analyses of the relationships between the dependent variable (Is a particular mushroom poisonous? "Yes," or "No") and 23 explanatory variables. Note that the dependent variable for this year's contest is qualitative, as are all 23 explanatory variables. Because there is not a quantitative variable in the entire data set, this problem is really not appropriate for students studying statistics for business and economics.

All 15 member colleges and universities in The Associated Colleges of the South require at least one statistics course of their majors in economics and/or business, and most member institutions offer and recommend a second course in statistics, econometrics, or quantitative models. I propose to offer a competition that will provide ACS professors an opportunity to involve their students of economics and business in the statistical analysis of a significant, real world data set under conditions that will simulate expectations placed on consulting statisticians.

An entry in the competition will consist of two parts: 1) ex post predictions of a dependent variable where the contest judges know the correct values, and 2) a written explanation of the prediction process (not to exceed 5 pages, including graphs). In other words, teams must justify their chosen model, just as a consulting statistician has to justify his or her work to the satisfaction of a client.

As a first pass through the entries, judges will determine which models have done the best job of predicting the (known) dependent variable values. Judgements will be based on a quantitative criterion such as mean absolute deviation. The best models will qualify for a second look by the judges: written reports will be evaluated on factors such as creativity, neatness, conciseness, and knowledge of methodology. The team that specifies a model that does a good job of predicting, and that explains their model in a compact, meaty presentation, will be declared the winner.

Each college may enter up to five teams in the competition with at most two students per team. A faculty advisor must sign each team's entry form, and a faculty advisor may be associated with more than one team. Advice from faculty should be very limited so that reports and predictions reflect the students' work. The role of the faculty advisor is to motivate, to keep their students focused, and to offer selected advice that will enhance the learning experience of the students. As a rule of thumb, if an advisor would not be willing to give this advice if the question arose as part of a take-home exam, the advice should not be offered. Contest judges will be recruited from among participating advisors at ACS schools.

An explanation of the contest, contest rules, and the problem will be published on a web server at Washington and Lee University early in the fall term of 1998. The data set to be analyzed will be made available for downloading at the same web site, as will the data set containing the relevant data for ex post predictions.

Student teams will submit completed registration forms and contest entries by uploading them to the server at W&L. Requests for clarifications of the rules or other matters will be asked and answered by means of e-mail.

Deadline for submission of the entry form, the written report, and ex post predictions will be February 1, 1999. Advisors will be notified of results by April 1, 1999. Winners will be announced on the contest web site, and appropriate press releases will be issued. Counting the number of colleges that participated, the number of advisors, and the number of entries will assess the project.

Professor Cline's workload will be tailored to compensate for time spent on creating and administering the competition by ensuring that he will have identical course preparations during the fall and spring terms.