Bias and Efficiency in Structural Equation Modeling: Maximum Likelihood Versus Robust Methods.

摘要: In the structural equation modeling literature, normal-distribution-based maximum likelihood (ML) method is most widely used, partly because resulting estimator claimed to be asymptotically unbiased and efficient. However, this may not hold when data deviate from normal distribution. Outlying cases or nonnormally distributed data, in practice, can make ML (MLE) biased inefficient. addition ML, robust methods have also been developed, which are designed minimize effects of outlying cases. But properties estimates their standard errors (SEs) never systematically studied. This article studies two compares them against with respect bias efficiency using a confirmatory factor model. Simulation results show that lead comparable normally distributed. When heavy tails cases, less more efficient est...