Structural Equation Models with Observed Variables
Kenneth A. Bollen · 1989
This chapter examines structural equation models with observed variables. First, they are the most common structural equation models. Second, these models are a special case of the more general structural equation procedures with latent variables that are discussed in the chapter. The major topics of the chapter-model specification, the implied covariance matrix, identification, and estimation—will recur for the other models. The maximum likelihood (ML) and generalized least squares (GLS) ones are asymptotically efficient when the assumption of multinormality holds or when the distribution of the variables have normal kurtosis, whereas the unweighted least squares (ULS) generally is inefficient. The chapter discusses several other topics that arise when utilizing observed variable models. These are standardized and unstandardized coefficients, alternative assumptions for x, interaction terms, and equations with intercepts.