An Investigation of the Reliability and Standard Error of Measurement of Words Read Correctly Per Minute Using Curriculum-Based Measurement

Abstract

Generalizability (G) theory was used with a sample of 37 third-grade students to assess the variability in words correct per minute (WCPM) scores caused by student skill and pas sage variability. Reliability-like coefficients and the SEM based on a specific number of assess ments using different combinations of passages demonstrated how manipulating probe vari ability could reduce measurement error. Results showed that 81 % of the variance was due to student skill, 10% was due to passage or probe variability, and 9% was due to unac counted sources of error. Reliability-like coeffi cients ranged from .81 to .99, and SEMs ranged from 18 to 4 WCPM depending on the number of probes given. When passage vari ability was controlled, SEMs were decreased and ranged from 12 to 4 WCPM. Results indi cated that WCPM scores yield high reliability- like coefficients but also have a large SEM that can be reduced by administering multiple alternate passages. Discussion focuses on con ducting research designed to identify more equivalent passages in order to reduce erro neous relative and absolute decisions.