Lastly, we compared WLW and frailty model (model (1.10)) results. Figure 3 shows the findings of an analysis involving the incident detection of 7 common oncogenic HPV, namely, HPV16/18/31/33/35/45/58 selected since the frailty model had problem converging when less common HPV types were analyzed (most likely because of insufficient occurrences to define their baseline hazards). Interestingly, unlike the comparison of mixed effects and GEE models, effect estimates obtained using the frailty model were only slightly larger than those obtained using WLW. Frailty models to assess type-specific associations failed to converge.
