Book Description
Many of the statistical techniques that have been used for developing and evaluating norm-referenced tests are not applicable to criterion-referenced tests (CRTs) and domain-referenced tests (DRTs) since the data from these later tests do not usually follow the normal distribution. Further, CRTs and DRTs are not used to compare or rank students against one another; rather, they are used to determine whether students have met or exceeded mastery learning levels or absolute performance standards. Statistical procedures are needed that can be easily employed by developers and evaluators of CRTs and DRTs in the Navy. The purpose of this effort was to investigate errors of measurement in criterion-referenced, domain-referenced, and mastery testing. A handbook of some statistical techniques for producing and evaluating DRTs was created for Navy practitioners. This is a 'how-to-do-it' guide for the intelligent layman who develops and assesses DRTs and/or CRTs. This handbook considers item analysis procedures, techniques for establishing cutting scores, errors of measurement and clssification, test length, and advancement scores, as well as group-based coefficients of agreement.