|
Abstract:
|
A recent study of vaginal HIV shedding followed HIV-positive patients with trichomonas matched to HIV-positive trichomonas-free patients for three visits. Vaginal shedding of the HIV virus was measured at baseline, at one month (post-treatment for the trichomonas patients), and at three months. The response distribution at each time point was a mixture of a proportion of zeros and a highly skewed distribution for those with detectable shedding. The analytical challenge was how to analyze these data to address the question of whether trichomonas increases vaginal HIV shedding, given that the analysis must incorporate the matching scheme, the longitudinal aspects of the data, and the mixture distribution of responses. Two-part models, logistic regression approaches, and other strategies were compared in terms of point estimators, variability, assumptions, and hypothesis testing.
|
