Confirmatory Factor Analysis and Reliability: Testing Measurement Model Assumptions

This paper demonstrates the use of confirmatory factor analysis by LISREL as an assumption-testing method when computing reliability coefficients. Formulae based on the LISREL estimates are presented for computing the coefficients under different measurement model assumptions. According to the empirical results, the reliability estimates obtained under the assumption of a well-fitting congeneric model were slightly higher than those obtained under the assumption of parallellity. However, the differences were small, an outcome which indicates that the reliability estimates are robust against departure from the assumption of parallellity of test items.