Structural Equation Modeling
Note: Presently, the German version of this page is far more comprehensive and up to date.
English versions of the introductory papers will be added and updated as time permits.
(1-8 pages each)
- Kline, R. B. (2010). Principles and practice of structural equation modeling. New York, NY: Guilford.
(accessible, up-to-date introduction to structural equation modeling)
- Bollen, K. A. (1989). Structural equation modeling with latent variables. New York, NY: Wiley.
(still one of the most comprehensive resources on structural equation modeling)
For the free statistics environment R (r-project.org), there are several packages for structural equation modeling:
(latent variable analysis)
by Yves Rosseel, Ghent University (Belgium):
lavaan is the newest structural equation modeling package for R. It employs an easy to learn, straightforward syntax for model definition by specifying individual paths, similar to general R conventions for model specification, and implements useful defaults (similar to the commercial Mplus software). The documentation is very easy to follow. lavaan already allows to analyze more complex models (particularly user-friendly specification of group comparisons) and can use different estimators (ML, GLS, WLS). new
sem package by John Fox, McMaster University (Hamilton, Ontario, Canada):
Historically the first structural equation modeling package for R, it allows ML and 2SLS estimation of models specified using individual paths. It is well documented, but presently does not offer more complex functionality (e.g., no group comparisons).
OpenMx re-implements the original free Mx software for structural equation modeling in R, offering comprehensive functionality and flexible model specification (based on matrices or based on paths between individual variables). It is however based on a rather complex syntax that does not necessarily follow general R conventions for model specification, or for the naming of objects. The documentation is comprehensive, but corresponding to the interface, complex.
For some commercial structural equation modeling packages, test/demo/student versions are available. The LISREL student edition allows relatively comprehensive analyses:
LISREL student edition for Windows;
the Linux version can currently be found here: ftp://ftp.lisrel.com/lisrel/linux/student/
(for course usage; allows to analyze models with up to 15 manifest variables)