A mixed effects model(18,19) is an alternative approach for incorporating repeated HPV prevalence data across HPV types and across visits involving the same woman. A mixed effects model for repeated assessments of HPV16 infection can be expressed as follows: (1.5)logitPij=μ+αi+δZij+ηWij, which allows each subject to have her own baseline rate of infection modeled by µ + αi where µ is the population average baseline rate of infection and αi represents the ith subject’s departure from the population average and is modeled as a random effect. The random effect is unobserved and characterized by a distribution function (typically a normal distribution centered at 0 with variation σ2), while Z and W are observed covariates and referred to as fixed effects. Model (1.5) is therefore called a mixed effects model. The parameter δ has a different interpretation from β in GEE: it is the log OR of HPV 16 infection in a woman in a certain CD4+ / HIV RNA stratum compared with her risk if she were HIV-negative. Since δ describes the change in risk within a person were her risk
