Computerized adaptive math tests for elementary talent development selection

A computerized adaptive test was developed for placing fifth grade elementary school children in a middle school mathematics talent development program. Three study cohorts were used for test construction, validation, and piloting. The 240 items that comprised the item bank were based on 60 objectives teachers identified as critical for successful completion of a sixth grade math talent development program. The items were partitioned into eight forms (average a = .76) with 10 anchor items and equated using Item Response Theory (IRT). A factor analysis of the 10 anchor items suggested that the item bank was essentially unidimensional (first factor explained 41% of the total test variance). A comparison between sixth grade children enrolled in a math talent development program and those enrolled in regular math classes provided additional evidence of construct validity. The scores generated by the computerized adaptive test correlated with the end‐of‐semester percent scores given by the teachers as well or better than CTBS (math), math grades, and teacher ratings. Because the test compares favorably with other predictors, it shows potential for talent development selection.