Using curriculum-based measures to predict math performance on a statewide assessment

One hundred forty-six students in third, fourth, and fifth grades completed two types of curriculum-based measurement (CBM) math probes, math computation and math concepts/ applications, in the winter and spring. The relationship between scores on the math probes and scores on the math portion of the Louisiana statewide assessments were analyzed by computing correlation coefficients and conducting multiple regression analyses. Both types of probes were significantly correlated with test outcomes, yet the concepts/applications probes were stronger predictors of test performance. The diagnostic accuracy of the probes was determined using receiver operating characteristic (ROC) curve analyses, which established cut scores that are associated with passing the statewide test and revealed the concepts/applications probes have higher overall correct classification rates. Having the ability to predict performance on high-stakes tests enhances the utility of CBM and provides the opportunity for educators to intensify instruction for at-risk students before they experience failure