Model study of the structure through a conditional multivariate analysis focused on talent identification in handball players

• Main Authors: J.J. Fernández, Mª H. Vila, F. A. Rodríguez
• Format: Article
• Language: English
• Published: Asociación Española de Ciencias del Deporte 2010-09-01
• Series: European Journal of Human Movement
• Online Access: https://eurjhm.com/index.php/eurjhm/article/view/107

The aim of this investigation was to analyse different anthropometrical, physical fitness and training characteristics of young handball players of different age categories from a multidimensional perspective, in order to obtain statistically developed reference norms for various testing procedures, and to build multivariate models that could predict performance level at different age periods (Solanellas & Rodríguez 1996). 105 handball players aged 13-18 years participated in the study, selected among the best players of the Galician Handball Federation (Spain). They were grouped into three official age categories: 13-14 (INF), 15-16 (CAD) and 17-18 (JUV). The multidimensional evaluation procedures included: 1) a specific questionnaire to analyse their sport participation background and training status; 2) a complete anthropometrical evaluation, including body composition analysis, somatotyping, and sexual maturation rating; 3) the Eurofit test battery (Council of Europe 1988) to measure general physical fitness; and 4) a vertical jump test battery (SJ, CMJ, and Abalakov). Different multivariate models were developed using discriminant analysis techniques (stepwise selection) to discriminate between players who were selected or not selected to play with the Galician national team by a committee of federal coaches. Wilks’ λ, F values, canonical correlation, and percentage of correctly classified players, among other parameters, were calculated. The predictive capacity of multivariate models developed by discriminant analysis reached 95% or more of players correctly classified when all variables were included. Variables entering the predictive model using the first discriminant function varied for each age category group, and correct classification percentage significantly decreased at the oldest age category (JUV). The variables entering the multivariate models with highest predictive value were predominantly those derived from physical fitness and anthropometrical tests. Training level appeared only at the oldest age category group. From the results, we conclude that the best age for talent detection based on this type of multidisciplinary evaluation (sports background and training status questionnaire, anthropometry, and physical fitness comprehensive testing) seems to be 15-16 years of age (CAD category), when coordinative and cognitive factors probably begin to play an increasingly important role in handball performance. These results could be particularly helpful in talent detection and development in younger players. KEY WORDS: multivariate analysis, talent detection y handball.