An Empirical Investigation of Equating Stability in a Single and a Double Linkage Design with Small Sample Sizes Using Angoff Model IV
Empirical evidence is presented of the relative efficiency of two potential linkage plans to be used when equivalent test forms are being administered. Equating is a process by which scores on one form of a test are converted to scores on another form of the same test. A Monte Carlo study was conducted to examine equating stability and statistical bias in a single and double linkage plan in small samples. Small random samples of 25, 50, and 100 were drawn with replacement from archival test data files representing Form B (base form), Form N (next form) and Form C (current form) pseudo-populations. Test data from two teacher certification subject area tests, Art Education and Hearing Impaired, both K-12 were used. Using the Angoff Model IV non-equivalent linear equating model, an indirect link, a direct link, and the average of the two links, equating equations were computed for each pair of samples at each sample size per subject area examination. Stability of the equating plans was evaluated by calculating the bootstrap standard errors of equating. Results indicate that the direct linkage design is more stable across raw score points, equating bias for direct linkage is trivial, and equating bias is quite large for the indirect linkage design. The direct linkage design is recommended for use with small sample sizes. Two tables and 13 figures illustrate the analyses. (Contains 12 references.) (SLD)
Du Bose, P. & Kromrey, J.D. An Empirical Investigation of Equating Stability in a Single and a Double Linkage Design with Small Sample Sizes Using Angoff Model IV.