Structural equation modeling has a long history dating back to the development of path analysis by Sewall Wright (Wright, 1921). Path analysis is an algorithmic tool for deriving a set of predicted covariances between variables which may be connected with either regression (asymmetric, directional) or correlation (symmetric, non-directional) paths. The advent of high speed computers and high level programming languages in the 1960's, together with advances in statistical methodology (Jöreskog, 1967) led to the development of software for fitting models to observed covariance matrices by maximum likelihood. This procedure is now commonly known as structural equation modeling (SEM). Several extensions of this methodology have increased its scope: modeling of means as well as covariances (Sörbom, 1974); specifying certain paths as observed variables (known as definition variables in Mx) (Neale, 1998; Neale, Boker, Xie, & Maes, 2006)); and fitting finite mixture distributions (Eaves, Neale, & Maes, 1996; Everitt & Hand, 1981; McLachlan & Peel, 2000).
