not know how the repeated HPV outcomes are correlated with each other, but a useful feature of the GEE approach is that even if the working correlation is mis-specified, the estimate of β is still unbiased. Further, because a robust variance is used the confidence interval for β is correct for large sample size. The main benefit of properly specifying the correlation between repeated observations is that if the working correlation chosen is close to the actual one the model has greater efficiency(11,12). In practice, it is recommended that several working correlations be assessed. The one that produces the smallest standard error for the estimate of β is generally accepted to be the one that is most consistent to the true correlation. In many cases, though, the results will be similar across different correlation assumptions. Time-dependent covariates can be incorporated.
