Unformatted text preview: Edward Rooker ILRST 210 11:15-12:05 First Article Evaluation Full Citation Jim Albert, Patricia Williamson. (2001). Using model/data simulations to detect streakiness. The American Statistician, 55(1), 41-50. Retrieved September 20, 2007, from ABI/INFORM Global database. (Document ID: 68156604). Primary Goal of Study To prove the statistical existence of "streakiness" in baseball hitting and basketball shooting data. Group Evaluated The article looks at Mark McQuire's home run statistics from 1995-1999 and Javier Lopez's hitting statistics in the 1998 season. Also, the study looks at the shots made by a college basketball player shooting 100 consecutive shots after 20 different practices. Who is Excluded From Study The study did not look at anyone who was not a member of a collegiate or professional sports club. Also, the study only focused on 2 sports, baseball and basketball. Altogether the study only looked at 3 athletes. Although, the study is very limited in the number of "participants" (only 3 athletes), the number of cases for each person studied is large, either looking at a large number of at-bats over years or in a season, or a large number of shots over a series of 20 practices. How did they collect the data? The data collected for Mark McGuire and Javier Lopez were statistics collected from other journal articles and studies conducted to look at the streakiness of these two athletes performance. The information on basketball shooting was collected from an experiment. The experiment recorded the outcome of 100 consecutive 3-point shots made by a basketball player on after 20 separate practices. Case Unit When looking at the baseball players each case is a single baseball game played by player. In the experiment with the basketball shots each case is a practice in which the results of 100 shots were recorded. Number of Cases For Mark McGuire, the study looked at 698 cases (games) over a 5 year period. For Javier Lopez the study looked at 132 cases (games) during the 1998 season. For the basketball experiment with the college basketball player there were a total of 20 cases (practices). Variables Collected For the Mark McGuire cases the study looked at the number home runs hit during each week of the five seasons, which is quantitative. For Javier Lopez the study looked at both quantitative and categorical data. The study looked both at Lopez's batting average each game (quantitative), as well as defining his performance by labeling each game as either "hot" or "cold" (categorical). For the basketball experiment the variable collected was whether or not each shot was a make or a miss (categorical). Major Variable Results For Mark McGuire Statistics Mean: 5.69 Standard Deviation: 1.09 For Javier Lopez Statistics Range of moving averages: .302 Sum of abs. diff. of moving averages: 6.27 Number of runs: 22 Length of longest run: 18 Logistic regression slope (subgroup 5): -.861 Logistic regression slope (subgroup 10): -1.217 Standard deviation of subgroup batting avg (subgroup 5): 0.0472 Standard deviation of subgroup batting avg (subgroup 10): 0.1005 For Basketball Shots The probability that the true slope of the regression will exceed 0 is 93%. Identify Statistical Tests or Measures The Mark McGuire study used a Beta distribution to look at streakiness. The study of Javier Lopez's statistics used the Markov switching model and a sequence of independent Bernoulli trials. The basketball shooting experiment used the Markov switching model to look at streakiness. Unmeasured Variables The study's final conclusion lacks certainty due to the fact that only 3 specific people were studied in 2 different sports. Possibly with further studies into multiple sports and many more athletes a more accurate conclusion on statistical proof in favor of "streakiness" in sports can be found. Also, the experiments were unable to keep all variables constant so outside forces may have effected the athletes' performance. Summary: This experiment attempted to find statistical proof that what is known in sports as "streakiness" existed and could be defined in a model. This statistical study looked at data collected from Mark McGuire's 1995-1999 seasons, Javier Lopez's 1998 season, and an experiment with a college basketball player shooting 100 consecutive shots after 20 practices. In the Mark McGuire study the major variables were the mean and standard deviation as well as the results of the beta distribution. When looking at Mark McGuire's statistics, his pattern of home run hitting was indistinguishable from the coin-toss model, meaning no "streakiness" could be determined. In Javier Lopez's study there were a number of important variables (listed above), but none of which "appear to provide support to the belief that Lopez was a streak hitter during this season." Lastly, the basketball experiment used a regression to interpret the data collected and found that the shooter's performance only improved with more practice, but no streakiness could be determined. The article's main fault could lie in the small number of individuals studied, even though the number of cases for each individual was sufficient. Overall the baseball data did not differ much from a coin-toss model and although the basketball experiment data did deviate from this model it still provided no clear statistical evidence for the statistical existence of streakiness. With more studies looking at a larger number of players and sports the conclusion found in this article would be more complete and may provide more statistical evidence for streakiness. ...
View Full Document
This note was uploaded on 12/13/2007 for the course ILRST 2100 taught by Professor Vellemanp during the Fall '07 term at Cornell.
- Fall '07