Using Evidence-Based Decision Trees Instead of Formulas to Identify At-Risk Readers
The purpose of this study was to examine whether the early identification of students who are at-risk for reading comprehension difficulties is improved using logistic regression or classification and regression tree (CART). This research question was motivated by state education leaders' interest in maintaining high classification accuracy while simultaneously improving practitioner understanding of the rules by which students are identified as at-risk or not at-risk readers. Logistic regression and CART were compared using data on a sample of grades 1 and 2 Florida public school students who participated in both interim assessments and an end-of-the year summative assessment during the 2012/13 academic year. Grade-level analyses were conducted and comparisons between methods were based on traditional measures of diagnostic accuracy, including sensitivity (i.e., proportion of true positives), specificity (proportion of true negatives), positive and negative predictive power, overall correct classification, and the receiver operating characteristic area under the curve. Results indicate that CART is comparable to logistic regression, with the results of both methods yielding negative predictive power greater than the recommended standard of .90. The comparability of results suggests that CART should be used due to its ease in interpretation by practitioners. In addition, CART holds several technical advantages over logistic regression.