Figure 2.

The power to detect an odds ratio of 2.0 for the association between prevalent HPV detection and host immune status, comparing HIV-positive women with CD4+ count <200 (n=535) versus >500 cells/mm³ (n=458) under various conditions: (i) using data from only a single observation (J=1; equivalent to standard logistic regression), (ii) using data from 5 or 10 repeated study visits per observations (J=5, J=10), and (iii) assuming the correlation between repeated observations is moderate (corr=0.3) or very high (corr=0.6). The figure shows greater statistical power for models that incorporate a generalized estimating equation (GEE), and use data from multiple observations, compared with standard logistic regression models, which can only use data from a single observation.)