Background:The preformance of triathletes can be predicted by the use of multivariate data analysis. It could be used to detect young talents or to redefine the training program of
Methods:To arrive at the optimum sample size G * Power software was used. 21 volunteers were selected, 7 professionals (group 1) and 14 amateurs (group 2), all male.
The variables used were the body mass index (BMI), age (A), resting heart rate (RHR), the number of years of practice of triathlon (YPT), the maximum oxygen consumption (VO2Max) , the weekly distance
training in swimming (WDS), the weekly distance training in cycling (WDC), the weekly distance training in running (WDR). The technique used was the discriminant analysis and the values were
normalized with natural logarithm (nl). All statistical analyze was performed using SPSS 21.
Results:The method used was stepwise, so that only in the model the variables with the greatest predictive power, considered jointly, not individually. The first group centroid
(professionals) stood at 2,338, while the second group centroid stood at -1,169. Making a weighted average of the centroid of each group, the sample size in each group, one reaches the Z cutoff score.
In this case, Z cutoff score would be zero. Above zero would be closer to the performance amateur athletes and values below zero would be closer to the performance of professional athletes. As the
sample was reduced was decided to work with the original models and cross-validation. The hit ratio stood at 92,93% in the case of the model with cross-validation.
Conclusions:This study showed that it's possible infer the triathletes performance, which is vitally important, whether for detecting talents, or for structuring the training.