Generalizability Theory and Many-Facet Rasch Measurement.
John Michael Linacre · 1993
Generalizability theory (G-Theory) and many-facet Rasch measurement (Rasch) manage the variability inherent when raters rate examinees on test items. The purpose of G-Theory is to estimate test reliability in a raw score metric. Unadjusted examinee raw scores are reported as measures. A variance component is estimated for the examinee distibution. Other variance components, due to item and rater distributions, interaction effects and random noise, are accumulated as examinee score error. Rasch computes a measure for each examinee, adjusted for the particular items and raters met by that examinee, that is more fair than the raw score. Rasch test reliability is higher than G-Theory reliability because Rasch error variance excludes item and judge variance.