Generalizability theory in the early stages of the observational method applied in the field of sports initiation: Data quality and sample size estimation

Abstract

This paper addresses, from the observational methodology, the study of early stages of sports initiation, in terms of reliability and esti- mation of the sample, from the use of the Generalizability Theory. The purpose of this study was: 1) to analyse data quality of the ad hoc observa- tional tool applied to Basque pelota, 2) to learn about the accuracy of generalizing from a hypothetical coding and recording of different numbers of games. Fleiss’ and Cohen’s Kappa coefficients were calculated, and an analysis of variance and generalizability were implemented. The results show that the observational tool allowed us to obtain a reliable record once the observers’ training process had been performed (Kappas above .80 and vari- ability of the observers close to 0) and, secondly, to estimate the level of accuracy generalization (generalizability coefficient, ξρ2(Δ)) from the coding of 4 (ξρ2(Δ) = .87), 6 (ξρ2(Δ) = .91), 8 (ξρ2(Δ) = . 93), 10 (ξρ2(Δ) = .95) or 15 (ξρ2(Δ) = .96) matches. Study results support the conclusion that we have a tool for evaluating the game that is applicable in the formative stages of Basque pelota, and the number of games necessary to generalize with a particular degree of accuracy.