Assessing the Classification Accuracy of Early Numeracy Curriculum-Based Measures Using Receiver Operating Characteristic Curve Analysis

Abstract

Receiver operating characteristic curve (ROC) analysis was used to investigate the ability of early numeracy curriculum-based measures (EN-CBM) administered in preschool to predict performance below the 25th and 40th percentiles on a quantity discrimination measure in kindergarten. Areas under the curve derived from a sample of 279 students ranged from .65 to .82. Optimal cut scores were associated with sensitivity ranging from .44 to .79, specificity ranging from .49 to .91, and classification accuracy ranging from .60 to .79. Among a validation sample of 140 students, sensitivity ranged from .36 to .83, specificity ranged from .46 to .89, and classification accuracy ranged from .57 to .79. Results indicate that EN-CBM administered in preschool predicts math performance in kindergarten and can be a valuable component of an evidence-based decision-making process.