Exploratory Data Analysis
Albert F. Smith State University of New York at Binghamton Deborah A. Prentice Princeton University
Top professional athletes are paid lavish sums of money for their services as members of sports teams. As each professional sports season begins, the headline sports story is frequently about the salary that has been negotiated between a star player and a team. Given the enormous financial investment a team makes in its players, it is reasonable to ask whether the investment pays off-that is, whether team success is related to the earnings of team members. As part of an inconclusive article that addressed this question, the New York Times published a display, from which Fig. 12. 1 is adapted, as one-third of its front sports page ( Bondy, 1991). For each of three professional leagues, teams are rank-ordered by number of salary dollars per victory achieved during the 1990-1991 seasons. This display is sadly uninformative in several respects, and we begin this chapter on the exploratory analysis of data by pursuing the question of how players' salaries relate to team success.
The most obvious shortcoming of Fig. 12.1 as a source of information about the relationship of salaries and team success is that it provides no information about team success. Although a committed fan may remember how many victories were achieved by each National Football League team during the 1990 regular season, most readers of the newspaper will not. If the price per victory for a team is high, but the team won many games, the team's management may consider the money to have been well spent. If the price per victory is low and the team won many games, the management has gotten a great deal (and perhaps exploited its players). Clearly, information about number of victories is essential.
The second shortcoming of Fig. 12.1 is the salary measure it provides. Dollars per victory would be an informative measure about the relationship between salaries and team success if every team in a league had the same salary budget.