Introduction to Structural Equation Modeling with Latent Variables

Modeling Languages for Specifying Models

PROC CALIS provides several modeling languages to specify a model. Different modeling languages in PROC CALIS are signified by the main model specification statement used. In this chapter, you have seen examples of the FACTOR, LINEQS, LISMOD, MSTRUCT, PATH, and RAM modeling languages. Depending on your modeling philosophy and the type of the model, you can choose a modeling language that is most suitable for your application. For example, models specified using structural equations can be transcribed directly into the LINEQS statement. Models that are hypothesized using path diagrams can be described easily in the PATH or RAM statement. First-order confirmatory or exploratory factor models are most conveniently specified by using the FACTOR and MATRIX statements. Traditional LISREL models are supported through the LISMOD and MATRIX statements. Finally, patterned covariance and mean models can be specified directly by the MSTRUCT and MATRIX statements, or by the COVPATTERN= and MEANPATTERN= options.

For most applications in structural equation modeling, the PATH and LINEQS statements are the easiest to use. For testing the built-in covariance and mean patterns of PROC CALIS, the use of the COVPATTERN= and the MEANPATTERN= options are the most efficient. In other cases, the FACTOR, LISMOD, MSTRUCT, or RAM statement might be more suitable. For very general matrix model specifications, you can use the COSAN modeling language. See the COSAN statement and the section The COSAN Model in ChapterĀ 29: The CALIS Procedure, for details about the COSAN modeling language. See also the section Which Modeling Language? in ChapterĀ 29: The CALIS Procedure, for a more detailed discussion about the use of different modeling languages.