Joint Model For Longitudinal And Time-to-event Data And Their Applications

In practice,longitudinal data and survival data often appear at the same time.For example,in many medical studies,information about patients changing over time,such as blood pressure,is often collected,and people are also interested in the time of recovery or recurrence.Longitudinal data and survival data are often associated in some way,and the time of events may be related to the longitudinal trajectory.Single analysis of longitudinal data and survival data may lead to invalid or biased results,while the joint model of longitudinal data and survival data contains all the information and provides effective inference.A familiar example is in the clinical trials of AIDS,including the treatment task,physiological characteristics,viral status,such as CD4 count and viral RNA copy number,and the time of development into AIDS or death.The joint models of shared random effects have attracted more and more attention in the study of the relationship between longitudinal and temporal event data,but these models are very complex in calculation.This thesis first introduces the longitudinal data model and survival data model.Then we introduce the joint model of shared random effects and its estimation methods,including maximum likelihood estimation,EM algorithm and Bayesian estimation.Then the combined model is applied to the specific AIDS data to study the relationship between CD4 count and mortality,and predict the data out of the sample.