Effects of Ignoring Correlated Measurement Error in Structural Equation Models

Abstract

The implications of ignoring correlated error on parameter estimates in some simple structural equation models are examined. Ignoring correlated measurement error not only affected the acceptability of a model based on goodness-of-fit measures, but also affected the parameter estimation. Ignoring within-construct correlated error does not affect the estimates of the structural parameter but overestimates the associated measurement parameters. It is shown analytically and empirically that ignoring positive between-construct correlated error overestimates the structural parameter linking the two constructs. The effects become more pronounced with decreasing reliability in measurement.