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 factor level to change, it is often referred to as the “subject-specific effect”(20,21). In addition to δ and η, the variation in the random effect σ can also be estimated from the model. The variation in the random effect is a measure of the amount of heterogeneity (i.e., the level of intra-subject correlation) in the population, if σ =0, the model reduces to a fixed effects model. Under a normal assumption, the variation in the random effect can be interpreted as follow: 95% of the subjects in the population will not depart from the population mean μ by more than 1.96σ.
