Search

In 1991 the Measurement and Evaluation Center of the University of Texas at Austin was asked to develop a test for credit by examination in four lower division courses in Japanese. The test (in Japanese) was constructed from locally developed items provided by instructors of Japanese. The developed test consisted of 80 items distributed among grammar, listening, and reading sections, with the listening section recorded on tape by instructors. The initial version was administered to 166 students in 1 course and 66 in another. Items with poor psychometric properties were edited or removed and new items were introduced. For validation, the revised test was administered to 170 students in spring 1992 (2 second semester courses) and 231 students in fall 1992 (2 first semester courses). Correlation coefficients for all four sections were acceptable. After analyzing the data and meeting with the instructors, test developers set cut scores for credit in each course. These scores allow students to earn credit with grades of A, B, or C. Sixteen tables display student scores, and one figure illustrates the cut scores. (SLD)

An attempt was made to determine which item response theory (IRT) equating method results in the least amount of equating error or "scale drift" when equating scores across one or more test forms. An internal anchor test design was employed with five different test forms, each consisting of 30 items, 10 in common with the base test and 5 to 10 in common with one or more other forms. Simulated data were generated for each using the Rasch model. Using one form as the base test, each of the others was equated directly to the base test and equated through one or more others to the base test. Equating methods examined were: (1) concurrent calibration; (2) equating constant procedure; (3) major axis procedure; and (4) fixed bs procedure. When equating error was assessed, it was found that concurrent calibration resulted in the least amount of equating error overall. When concurrent calibration is not feasible, results indicate that major axis equating results in the least amount of equating error when equating across one or more forms. (Contains 5 references, 6 tables, and 1 figure.) (Author/SLD)

This study compared several item exposure control procedures for computerized adaptive test (CAT) systems based on a three-parameter logistic testlet response theory model (X. Wang, E. Bradlow, and H. Wainer, 2002) and G. Masters' (1982) partial credit model using real data from the Verbal Reasoning section of the Medical College Admission Test. Responses of an average of over 7,000 examinees for each of 22 test forms were used. The exposure control procedures studied were the modified within 0.10 logits procedure (L. Davis and B. Dodd, 2001), the Sympson-Hetter procedure (J. Sympson and R. Hetter, 1985) with a maximum exposure rate restricted to 0.19, the Sympson and Hetter procedure with a maximum exposure rate restricted to 0.29, and maximum information, a no exposure control condition, used as a baseline. The exposure control procedures were evaluated for measurement precision, utilization of the item pool, and item overlap across test administrations. For both measurement models, the modified within 0.10 logits procedure provided better pool utilization, with little decrement in precision of measurement, than either of the Sympson-Hetter procedures. (Contains 8 tables and 27 references.) (Author/SLD)

Human beings judge images by complex mental processes, whereas computing machines extract features. By reducing scaled human judgments and machine extracted features to a common metric space and fitting them by regression, the judgments of human experts rendered on a sample of images may be imposed on an image population to provide automatic classification.