[S] yags = yet another GEE solver

Vincent Carey 525-2265 (stvjc@channing.harvard.edu)
Fri, 10 Apr 98 17:15:12 EDT


source code and doc for yags, yet another GEE solver,
is available at

http://biosun1.harvard.edu/~carey/gee.html

The peculiar name is chosen to reduce the likelihood
of confusion with any other GEE solvers that might
be in use.

As noted in my previous posting, this is a slow
but very flexible implementation of GEE. Extensive
documentation is provided. The user interface
is quite similar to that of "gee()", but the "corstr"
parameter has a more limited range. A functionality
improvement relative to "gee()" is the use of
information on observation times in the AR(1) and
unstructured working correlation models, so that
gappy repeated measures can be accommodated with
these models.

Examples illustrate how to code for heterogeneous
correlation structures and how to use likelihood to
estimate correlation parameters. The documentation
on and coding methods for extending GEEs in this
framework still needs a lot of work. However,
extension and maintenance in this framework will
be much simpler than it is for "gee()", and there
is no memory leakage problem with yags(). It
is not necessary to have a C compiler to use
yags(), although a somewhat helpful patterned
matrix library is provided that will speed
certain calculations when loaded.

-- 
stvjc@gauss.bwh.harvard.edu
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu.  To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message:  unsubscribe s-news