Feel free to take it and use it: formal details in the file COPYING
I have tested it against S-PLUS and SPIDA but I don't guarantee it's bug-free. Take care.
Last updated February 5, 1997.
The December 12,1996 version is Issue 3, Volume 1 of the Journal of Statistical Software An earlier version of the system was also presented at a poster session at the Pacific Northwest Statistical Meeting, April 1996, where it won a student poster prize.
gee-modelor by setting the global variable
tVersion 1. Most of the useful working correlations included. The GEE system has been improved to use a simple model formula command for factor variables and interactions. This version requires the model formula code as well as the gee code. The model formula code also contains methods for use with regression-model-proto and functions for constructing design matrices for any purpose. Numerical robustness has been improved so that strange combinations of link and variance function can be fitted.
The formula objects don't work with version 3.47 of XLISP-Stat because it has a bug that prevents inverting 1x1 matrices.
Some diagnostics have been added, using the dynamic graphics and linking abilities of Lisp-Stat. These include various flavours of residual and a set of deletion diagnostics based on the work of John Preisser. I strongly recommend getting de Leeuw & Udina's gnuplot interface (code, documentation) for producing publication quality graphs.
Documentation is now available online for GEE . and model formula code
gee.1.0.ps or gee.1.0.tex GEE documentation
example.lsp karim.lspexamples and data used in the documentation
modelformula.lsp Model formula code
modelformula.txt or modelformula.ps Model formula documentation
geetest.lsp , testdata.dat, geetest.out, modelexamples.out: Test files (based on the Splus gee() library by Carey & Macdermott)
by.lsp is a function to perform analyses on subsets specified by the values of a set of variables. For example, you might want mean age stratified by sex and smoking status, or a separate gee-model for each state in your dataset . The function is fairly general but can't handle model formulas.