Estimating confirmatory factor and structural equation models (SEMs) usually requires specialized software such as Amos, LISREL, EQS, or Mplus. Each of these options has distinct advantages (Mplus in particular can estimate some really cool models), but unfortunately licenses can be very expensive. Free trial versions do exist for LISREL and Mplus. However, these put strong restrictions on the types of models that can be estimated.
In terms of general-purpose statistical packages, PASW (SPSS) cannot estimate SEMs directly; SPSS distributes Amos as a separate license. Those who already have a SAS license can take advantage of the CALIS procedure. Stata users can download the glamm program, which is capable of handling both latent variable and multilevel models.
A free alternative is to use one of the SEM packages written for R. The sem package, prepared by John Fox, can estimate straightforward confirmatory factor and structural equation models. This helpful paper introducing the sem package also includes examples of how one can take advantage of R’s many other capabilities to carry out more complicated estimation, such as latent variable models with ordered manifest variables. An example of a confirmatory factor analysis using sem can be found here. An example of estimating a structural equation model can be found here.
Yet another alternative is OpenMX, a newer library of R functions for estimating structural equation models. The OpenMx User’s Guide contains an introduction to the syntax needed to estimate latent variable models.
To install OpenMx, enter the following at R’s command prompt:
Once installed, it is possible to use OpenMx in any new R session by typing:
Still have questions? Contact us!