Comparing methodologies used to predict outcomes on high stakes tests with curriculum-based measurement
Brown, Sarah Eleanor
MetadataShow full item record
This dissertation investigated the ability of oral reading fluency (ORF) and ORF growth to predict outcomes on a high stakes achievement test. Predictive methodologies were used in order to predict outcomes using those measures. Data were collected regarding students’ ORF and their performance on a high stakes achievement test at a later date in time. ORF scores at different time periods were used to create an ORF growth score. ORF and ORF growth were used to predict outcomes on the high stakes test. Predictive methodologies of direct logistic regression and discriminant function analysis were used. It was found that ORF and ORF growth predicted outcomes at all grade levels with discriminant function analysis, but with logistic regression, only ORF or ORF growth predicted outcomes at some grade levels. Therefore, although further evidence of the ability of ORF to predict outcomes on high stakes tests was obtained, conclusive evidence of how ORF and ORF growth can predict outcomes using predictive methodologies was not found. Additionally, receiver operating characteristic (ROC) analysis was used to create cut scores specifically for this sample of students. Those cut scores were then used to predict outcomes on a high stakes test for those students. It was found that cut scores chosen specifically for this sample were higher than traditionally used cut scores, suggesting that schools may want to examine the predictable nature of the cut scores used in order to determine if they meet the needs of their specific populations.