General purpose packages providing GEE software include SPIDA (Statistical Laboratory, Macquarie University, Australia) and Stata (StataCorp). I am also informed by a reviewer that SAS 6.12 (SAS Corporation) includes this facility in PROC GENMOD but have not seen this version of the package. The implementation in SPIDA is somewhat limited, providing only the canonical link for the Normal, binomial and Poisson variance functions. The ordering of observations within a group is assumed to be correct in the data, leading to possible misleading results. Stata allows more freedom in combining link and variance functions but still prohibits, for example, binomial data with identity link. The Stata implementation requires a time identifier for correlation structures other than exchangeable and independence, but does not correctly handle missing observations. When the observation times for different groups are different a warning is produced to indicate that the results may be incorrect but the calculations are carried out as if the non-missing observations were actually adjacent in time.
User-contributed code is available for SAS and S. The GEE macro for SAS by Karim and Zeger and the cgee library for S by Carey and Macdermott provide similar functionality to Stata but require that the data are presorted by group and then by time within group. The S library does not check this and will return incorrect answers with no warning if the data are not correctly sorted.
While these programs (with the possible exception of the new SAS PROC GENMOD which I have not examined) provide fewer features than this Lisp-Stat implementation and in particular do not provide diagnostics they all have the advantage of computing environments providing superior data management and editing facilities to Lisp-Stat.