|
Abstract:
|
In HIV vaccine studies, a major research objective is to identify immune response biomarkers measured longitudinally that may be associated with risk of HIV infection. This objective can be assessed via joint modelling of longitudinal and survival data. Joint models for HIV vaccine data are complicated by the following issues: (i) some longitudinal biomarker data may be left censored due to lower limits of quantification, and distributional assumptions for censored values may be unreasonable; (ii) the longitudinal multivariate biomarker data are intercorrelated and may be of mixed types such as binary and continuous; and (iii) the longitudinal data may be measured with errors and have missing values. Moreover, the computation associated with likelihood inference can be highly demanding. In this paper, we propose a joint model and a computationally efficient method to address the foregoing issues simultaneously. In particular, our proposed method for censored longitudinal data does not make unverifiable distributional assumptions for censored values, which is different from methods commonly used in the literature. Data analysis and simulation study are also presented.
|
