Abstract:
|
In many longitudinal studies, subjects are at risk of experiencing an event that may substantially impact their response profiles. When an event is likely to alter subsequent measures of a serial response, a statistical analysis of the response profile should consider the risk of an event occurring. Motivated by a depression prevention study among patients with malignant melanoma, we examine a joint model that incorporates event risks into the analysis of serial depression measures. The joint model provides estimates of association between the longitudinal responses and event risks. We present a maximum-likelihood estimator for the mean response vector, conditional on given risk levels. The joint model permits hypothesis testing about functions of mean response profiles, as well as testing for association between the response and event risk. We illustrate the application of our joint model using the depression data, and we show benefits of the joint model over a standard mixed effects model.
|