a review of the meta-analytic studies of reliability generalization - una revisión de los estudios meta-analíticos de generalización de la fiabilidad

Abstract

The meta-analytic approach of reliabilitygeneralization (RG) pretends to show that reliability is anempirical property that varies from one test application toanother. This recent meta-analytic approach is helping tomake the researchers aware of the importance of reportingreliability estimates obtained from the own data and, ofavoiding the malpractice of inducting reliability coefficientsfrom other studies and previous applications of thetest. The stages to carry out an RG study are presented: (a)formulating the problem, (b) searching for the studies, (c)coding studies, and (d) statistical analysis and interpretation.An updated overview of the statistical problems ofthis approach is also offered: (a) to transform versus not totransform the reliability coefficients, (b) to weight versusnot to weight the coefficients, and (d) which statisticalmodel is the most appropriate (fixed-, random-, andmixed-effects). A systematic review of the 49 RG studiespublished to date is presented with the purpose of analyzingthe heterogeneity in how the data are statisticallyanalyzed. Finally, the implications of the RG studies forresearch and professional practice are discussed.