Improving the Reliability of Curriculum-Based Measures of Academic Skills for Psychoeducational Decision Making

Abstract

The effects of aggregation on the reliability of curriculum-based measuresof academic performance were explored in two studies. In the first , 30 children were tested four times, over four consecutive weeks, on three reading tests. Group stability coefficients, within-subject reliability coefficients, and group correlations between variables were calculated on the basis of one testing and then on the basis of aggregations over four testings. On the standardized measure and on the curriculum-based test correct score, reliability was high initially and aggregation had little impact. However, on the curriculum-based test correct score, reliability was high initially and aggregation substantially improved reliability, as evidenced in all three analyses. In the second study, 78 children were tested 10 times, over 10 weeks, on alternate forms of two curriculum-based reading measures and one curriculum-related written expression test. Group stability coefficients were calculated on the basis 2, 4, 6, 8 and 10 observations. For the reading test correct score, an initially precise measure, aggregation had little effect, whereas aggregating over occasions and test forms dramatically improved the reliability of the initially inaccurate measures, the reading test error score and the written expression score. Implications for the measurement of curriculum-based academic behavior are discussed.