GEE models, furthermore, can be used to incorporate data across multiple types of HPV. In a single GEE logistic regression model we can treat each of the several different oncogenic HPV types as separate endpoints, and then estimate the exposure effect on oncogenic HPV as a whole. We note that this common effect can also be examined with “any oncogenic HPV” as a single binary outcome. However, greater efficiency is obtained for grouped HPV data since each HPV type separately contributes data to the analysis, and the final result is a weighted average across HPV types. This source of efficiency is in addition to that discussed above, of being able to assess repeated visits over time.
