Evaluating Tau-U With Oral Reading Fluency Data and the Impact of Measurement Error.

Abstract

Recently, researchers have argued that using quantitative effect sizes in single-case design (SCD) research may facilitate the identification evidence-based practices. Indices to quantify nonoverlap are among the most common methods for quantifying treatment effects in SCD research. Tau-U represents a family of effect size indices that were developed to address criticisms of previously developed measures of nonoverlap. However, more research is necessary to determine the extent to which Tau-U successfully addresses proposed limitations of other nonoverlap methods. This study evaluated Tau-U effect sizes, derived from multiple-baseline designs, where researchers used curriculum-based measures of reading (CBM-R) to measure reading fluency. Specifically, we evaluated the distribution of the summary Tau-U statistic when applied to a large set of CBM-R data and assessed how the variability inherent in CBM-R data may influence the obtained Tau-U values. Findings suggest that the summary Tau-U statistic may be susceptible to ceiling effects. Moreover, the results provide initial evidence that error inherent in CBM-R scores may have a small but meaningful influence on the obtained effect sizes. Implications and recommendations for research and practice are discussed.