The method of generalized estimation equation (GEE), similar to ESS method, has a goal of utilizing correlated samples in an analysis (Liang and Zeger, 1986; Hanley et al., 2003). However, one major difference between GEE and ESS is that GEE relies on data to estimate the within-cluster correlation among samples, whereas ESS calculates the correlation by the information given. Typically in GEE, only a single correlation coefficient r is estimated for all clusters, which can be unreliable if clusters of different natures are included in the data. For example, if the dataset contains both sibpairs and cousin-pairs, the r for samples within sibpairs should be larger than that for cousin-pairs. Another difference is that GEE corrects not only variance, but also mean as well, whereas ESS only modifies variance. Similar to an argument made in (Devlin et al., 2004), we believe that sample correlation mainly affects the variance, and has less effect on bias.
