Efficiency of an Adaptive Inter-Subtest Branching Strategy in the Measurement of Classroom Achievement

摘要

A real-data simulation was conducted to investigate the efficacy of an adaptive testing strategy designed for achievement test batteries applied to a classroom achievement test. This testing strategy combined adaptive item selection routines both within and between the subtests of the test battery. Comparisons were made between the conventionally administered tests and the simulated adaptive tests in terms of test length, psychometric information, and correlations of achievement estimates. Design of the study also permitted (1) separation of the effects of the adaptive intra-subtest item selection procedure and inter-subtest branching, (2) evaluation of the effects of different intra-subtest termination criteria, (3) use of classical regression equations and regression equations corrected for errors of measurement in the predictors, and (4) cross-validation stability of the inter-subtest branching regression predictions. Data consisted of the responses from 1,600 students to classroom-administered final exams in a general biology course at the University of Minnesota. Results from this study generally supported the generality of this adaptive testing strategy for reducing achievement test length with no adverse impact on the quality of the measurements. Suggestions are made for further research with this testing strategy.