Application of the Generalizability Theory in sport to study the validity, reliability and estimation of samples

Abstract

The generalizability theory (G theory) is a multifaceted error theory that assumes that any measurement situation possesses infinite sources of variation. In the field of observation, it can therefore be applied to determine the influence of these sources of error on measurements. By way of an example, this paper describes three applications of the G theory in the initial stages of observational research: 1) to study validity; 2) to estimate the re- quired sample size; and 3) to study reliability. Currently available software applications (such as GT, EduG or, recently, SAGT) make it easier to conduct this kind of analysis and they provide researchers with procedures for ensuring that the decisions they need to make during the research process are justified prior to being implemented; for instance, calculating the required sample size so as to ensure the accurate generalization of results, or constructing a valid reliable observational instrument that can meet the arduous task of coding and recording behaviour in natural contexts.