Measuring and Predicting Pupils’ Progress in Special Education

Abstract

In regular primary education, LVS tests are used to measure and predict pupils' progress in math ability. In special education, due to measurement error and a distribution of scores that is different from regular education, both measuring and predicting progress is diffcult. A solution to the problem in measurement is investigated in the field of computerized adaptive testing (CAT). Using a simulation study, two CAT item selection mechanisms have been tested, for three different ability levels. Results suggest that using CAT mechanisms, accurately estimating a pupil's ability is possible, as long as the items are suitable for the ability levels. Overall, allthough the selection mechanisms are different, both mechanisms show equal performance in accuracy of measurment. For the problem of predicting progress, a distribution for special education is calculated and tested. The results suggest that, due to regression to the mean, the special education distribution does not predict accurately for pupils with values far from the population mean. Using Growth Mixture Modeling (GMM) multiple distributions have been defined to solve this problem. Using multiple distributions for pupils with different ability levels, the problem of regression to the mean is less severe.