The Application Of Discriminant Analysis And Logistical Regression as Methods Of Compilation In The Prediction Function In Youth Rugby

Por: Conrad Booysen e Pieter Ernst Kruger.

Athens 2004: Pre-olympic Congress

Send to Kindle


Introduction
Talent identification is a process where potential sportsmen/women are identified and developed in a specific type of
sport. One of the primary reasons for talent identification is to ensure the future survival and development of the
particular sport in which the talent is identified and sought. Worldwide there is a move toward the identification of
talent, with certain countries ahead in the process, successfully applying it to their national sporting codes. Due to our
exclusion from international competition for many decades, we find ourselves in the position of being behind in the
field of talent identification, albeit catching up very fast. What has been noticed however, is the move towards the
favoring of two primary methods of talent identification, namely discriminant analysis and logistical regression.
Different countries of the world use either of these methods, with an open debate currently being fought as to the
predictive abilities of these respective methods.
Aim
This study compares the two main models of talent identification i.e.: discriminant analysis and logistical regression in
terms of their ability to predict talent and their prediction functions

Method
Two groups of rugby players were identified and tested. They were the SARFU u/12 group (n=43) and the North West
(MSP) u/12 group (n=40). The motor/physical abilities tests that were adminstered were sprint time/speed, agility run,
flexed arm hang, vertical jump and speed endurance, 18 anthropometrical tests were conducted, and the original rugby
skills tests included passing for distance, passing for accuracy over 4m and 7m, running and catching, kicking for
distance and kick-off for distance.

Results
The data was analysed according to the discriminant analysis method and the logistical regression method. The results
were then compared according to the respective predictions of both methods

Discussion/Conclusion
The finding of this study is there are no significant differences found with the results obtained by using discriminant
analysis and logistical regression as methods of predictive function (Table 1). Each method has a 100% accuracy in
predicting or distinguishing between more and less talented players. Based on this it would then be up to the
discrimination of the researcher(s) as to which method would be used. Discriminant analysis provides an accurate view
of the best discriminating factors involved in talent, while logistical regression provides a view of the relative impact of
the various factors that determine talent. As can be seen, both of these models can and should be used effectively in the
identification of talent. Through various methods of statistical evaluation and substitutions, it is found that by
combining the two models a less accurate predictor is formulated, negating the need to combine the two models. This
then elicited the conclusion that each model, whilst both being 100% accurate, could be applied under different
circumstances when different information is sought. The proposition is made that when talent is identified, both models
be used rather than following an "either-or" approach. The advantage with this is that a well-rounded view of the
individuals under scrutiny after applying both models is formulated. Therefore, in summary it can be said that both
models predict with the same accuracy, with each model having it’s own unique areas of application.

NOTA: O texto com a iconografia está no anexo.

Ver Arquivo (PDF)

Comentários


:-)





© 1996-2019 Centro Esportivo Virtual - CEV.
O material veiculado neste site poderá ser livremente distribuído para fins não comerciais, segundo os termos da licença da Creative Commons.