Does Greatest Lower Bound Outperform Coefficient Alpha Under Violation of Essential Tau-Equivalence?

• In Event: Poster Session 4
  In Poster Session: Division D Section 1 Poster Session

Abstract

Coefficient alpha is the most frequently used and reported index of reliability because of its simplicity in computation and accessibility in statistical programs. However, coefficient alpha has been frequently criticized for being negatively biased when essential tau-equivalence is violated. This study is designed to use simulated data to evaluate whether the greatest lower bound (glb), as an alternative to coefficient alpha, can improve the accuracy of reliability estimation. In light of our simulation results, we recommend considering the glb as an alternative to estimating reliability only when a test has strong group factors, a relatively small number of items (20 or less), and a large sample size (500 or more).

Authors

• Wei Tang, University of Alberta

• Ying Cui, University of Alberta