Walter R. Mebane Jr. - Jasjeet S. Sekhon
Date: August 25, 1998
Professor Jasjeet S. Sekhon
Professor Walter R. Mebane Jr.
Please see http://data.fas.harvard.edu/jsekhon/ for the latest binaries, documentation and links to related academic articles. Direct all questions and comments to Jasjeet Sekhon.
GENBLIS (GENetic optimization and Bootstrapping of LInear Structures) is an evolutionary program which estimates structural equation models with latent variables by combining the global search strength of evolutionary algorithms with the ability of derivative based algorithms to do rapid local hill climbing. GENBLIS provides bootstrap confidence intervals for parameters and bootstrap confidence limits for the goodness-of-fit test statistic. Structural equation models are often used in political science, sociology, psychology, marketing and related social science fields to specify and estimate stochastic models for linear relations in the presence of measurement errors.
There are a variety of ways to parameterize structural equation models.
Following Bollen (1989), GENBLIS uses a
parameterization developed by Jöreskog (1973)
and Wiley (1973).
This parameterization has become a standard because of the popularity of
the LISREL program which was developed by Jöreskog and Sörbom.
Bollen (1989) gives an introduction to LISREL
models, including good review of
theoretical and applied literature up through the mid-1980s.
Small parts of the current GENBLIS code were adapted from an early version of GENOCOP I. For information about GENOCOP III please see http://www.coe.uncc.edu/~gnazhiya/gchome.html. You may also email its primary author Zbigniew Michalewicz.