|
17th
Annual ECSS-Congress, Bruges 2012
Abstract details
| Abstract-ID: |
1384 |
|
Session: |
[PP-SH23] Sport Statistics and Analyses [SA] 2 |
|
Lecture room: |
Sala professorat 2 |
|
Date & time: |
28.06.2013 / 15:00 - 16:00 |
| Title of the paper: |
MULTIVARIATE STATISTICAL APPROACHES TO TALENT IDENTIFICATION |
| Authors: |
Heazlewood, I. |
| Institution: |
Charles Darwin University |
| Department: |
Exercise and Sport Science |
| Country: |
Australia |
|
Abstract text |
Do not insert authors here
Introduction
Talent identification (ID) is a major focus of many national sports associations. Sport promotes national pride, such as World Championships, Olympic and Commonwealth Games Paralympic Games. The foci of talent ID is to identify talent for World Championships, Olympic, Commonwealth and Paralympic Games; identify human talent for professional sports due to high turnover of athletes; enhance the use of limited resources by targeting the appropriate athletes and identify the next generation of high performance athletes. Many programs in Australia and oversees are directed towards target populations, such as children, adolescents and adults athletes.
Predictive problems with univariate statistical approaches to talent ID include analysis of variables in isolation and they do not describe complex interactions that often occur between many tests that are conducted in talent ID. A significant level of redundancy within test variable sets may exist and this redundancy is difficult to evaluate using the univariate approaches. Multivariate prediction models permit a more holistic analysis of factors (Arbuckle, 2009; Hair et al., 2010) that explain and predict high performance athletes. Multivariate methods exist that can explain complex interactions that predict competition performance (Vaeyens, 2008).
Methods
A number of sports were analysed using multivariate statistical methods, such as multiple linear regression using kinanthropometric, exercise physiological, biomechanical and sports psychological factors to predict high performance triathlete times, factor analysis to examine the more complex relationship between torque, power, work, fatigue and acceleration, path analysis and structural equation modeling to predict performance outcomes in the decathlon and heptathlon, kinanthropometric, exercise physiological, biomechanical factors predicting BMX competition performance and neural network analysis to classify karate ability based on general motor and specific motor fitness tests.
Results
The multivariate methods were capable of developing hierarchies of importance for the predictor variable sets that explain and predict competitive performance. The explained variance utilising multivariate statistical methods was more substantive and enabled redundancy analysis and the identification of critical predictive subsets of factors predicting performance.
Discussion
Multivariate statistical approaches to talent identification provide both more meaningful and more complex explanations in predicting performance and as a more complete method of talent identification.
References
Arbuckle J (2009). AMOS 18 Users Guide. SPSS Inc, Chicago.
Hair JE, Black W, Babin BJ, Anderson, RE (2010). Multivariate Data Analysis (7th Ed.). Pearson - Prentice Hall, Upper Saddle River.
Vaeyens R, Lenoir M, Williams A, Philippaerts R (2008). Sports Med. 38(9), 703-714.
|
| Topic: |
Sport Statistics and Analyses |
| Keyword I: |
multivariate |
| Keyword II: |
talentID |
| Keyword III: |
sport |
|