Blackwell Reference Online

Extract

Reliability generalization (RG) is a measurement meta-analytic method proposed by Vacha-Haase (1998) characterizing score reliability across multiple administrations of a measure as well as identification of study features predictive of measurement error variation. RG identifies (1) the typical score reliability for a given measure, (2) the variability in score reliabilities across administrations of a given measure, and (3) which features of the measurement protocol do and do not explain or predict these variations in score reliability. Reliability generalization is similar to the theoretical concepts of validity generalization. Meta-analysis of validity coefficients were the precursors to validity generalization, which began in the late 1970s to test whether the validity of scores for a given measure or set of related measures was generalizable. Meta-analysis was the testing of the hypotheses of “situation-specific validity.” In validity generalization inquiries, studies are used as the unit of analysis, and means, standard deviations, and other descriptive statistics are computed for the validity coefficients across studies. The validity coefficients across studies may also be used as the dependent variables in regression or other analyses. In these analyses the features of the studies (e.g., sample sizes, types of samples, ages of participants) that best predict the variations ... log in or subscribe to read full text