A search of the PsycInfo database for “latent variable” or “latent class” or “structural equation model” gives an estimate of the increasing popularity of SEM: 1970's, 23; 1980's, 357; 1990's, 2,794; and 2000-2009, 9,599. These searches underestimate the actual number of published articles that used this method, as many abstracts do not provide detail about the statistical methods used. At the same time, the great variety of statistical methods that are subsumed by SEM including analysis of variance, multiple regression, discriminant analysis, canonical and partial correlations, factor analysis, principal components analysis and multilevel analysis, (Marcoulides & Schumacker, 1996; McArdle & Hamagami, 1996; Longford & Muthén, 1992) further demonstrate its broad utility.
