Structural Equation Modeling Estimates of Reliability for Ordered Categorical Items: Omega Versus Nonlinear Reliability Coefficients

• In Event: Poster Session 15
  In Poster Session: Important Issues in Model Selection, Measurement Invariance, Reliability, and Mixture Models

Abstract

We evaluated nonlinear (NL) and omega coefficients based on structural equation modeling to assess the reliability of scales. Item frequently have ordered categorical (OC) response scales. NL coefficients take into account the nature of these items by calculating item thresholds, factoring polychoric correlations, and computing reliability coefficients based on the estimated parameters. In comparison, omega coefficients ignore that items are OC. We focused on bifactor models that specify a general factor and group factors for subsets of items. With bifactor models, overall omega and hierarchical omega yield less accurate estimates of NL reliability coefficients when the thresholds are heterogeneous. In contrast to omega, the overall NL reliability does not decompose into reliabilities due to the general factor and group factors.

Authors

• Samuel B. Green, Arizona State University

• Yanyun Yang, Florida State University