The determinacy of variables in a structural equation model is considered, and it is shown that when the variables in the structural model are all manifest, the error-terms (disturbances) are uniquely determined given the parameters of the data. When the variables are all latent, the error-terms have indeterminate components. (SLD)

Structural equation modelling is becoming increasingly popular in education. This article examines and compares a number of alternative assumptions governing nondirected paths in structural equation models without latent variables vis-a-vis a data set on lung ventilation. Some problems with the conventional procedures in path analysis are pointed…

Conventional structural equation modeling fits a covariance structure implied by the equations of the model. This treatment of the model often gives misleading results because overall goodness of fit tests do not focus on the specific constraints implied by the model. An alternative treatment arising from Pearl's directed acyclic graph theory…

An algorithm for obtaining initial values for the minimization process in covariance structure analysis is developed that is more generally applicable for computing parameters connected to latent variables than the currently existing ones. The algorithm is formulated in terms of the RAM model but can be extended. (SLD)