''Preface Joint models for longitudinal and time-to-event data have become a valuable tool in the analysis of follow-up data. These models are applicable mainly in two settings: First, when focus is in the survival outcome and we wish to account for the effect of an endogenous time-dependent covariate measured with error, and second, when focus is in the longitudinal outcome and we wish to correct for nonrandom dropout. Due to their capability to provide valid inferences in settings where simpler statistical tools fail to do so, and their wide range of applications, the last 25 years have seen many advances in the joint modeling field. Even though interest and developments in joint models have been widespread, information about them has been equally scattered in articles, presenting recent advances in the field, and in book chapters in a few texts dedicated either to longitudinal or survival data analysis. However, no single monograph or text dedicated to this type of models seems to be available. The purpose in writing this book, therefore, is to provide an overview of the theory and application of joint models for longitudinal and survival data. In the literature two main frameworks have been proposed, namely the random effects joint model that uses latent variables to capture the associations between the two outcomes (Tsiatis and Davidian, 2004), and the marginal structural joint models based on G estimators (Robins et al., 1999, 2000). In this book we focus in the former. Both subfields of joint modeling, i.e., handling of endogenous time-varying covariates and nonrandom dropout, are equally covered and presented in real datasets''-- Read more...
Content: Introduction Inferential Objectives in Longitudinal Studies Case Studies Organization of the Book Analysis of Longitudinal Data Features of Repeated Measures Data Linear Mixed Effects Models Dropout in Longitudinal Studies Analysis of Time-to-Event Data Features of Event Time Data Relative Risk Models Time-Dependent Covariates Joint Models for Longitudinal and Time-to-Event Data The Standard Joint Model Connection with the Dropout Framework Extensions of the Standard Joint Model Parameterizations Multiple Failure Times Latent Class Joint Models Diagnostics Residuals for the Longitudinal Submodel Residuals for the Survival Submodel Random Effects Distribution Prediction and Accuracy in Joint Models Dynamic Predictions for the Survival and Longitudinal Outcomes Effect of the Parameterization on Predictions Prospective Accuracy Measures for Longitudinal Markers
Abstract: ''Preface Joint models for longitudinal and time-to-event data have become a valuable tool in the analysis of follow-up data. These models are applicable mainly in two settings: First, when focus is in the survival outcome and we wish to account for the effect of an endogenous time-dependent covariate measured with error, and second, when focus is in the longitudinal outcome and we wish to correct for nonrandom dropout. Due to their capability to provide valid inferences in settings where simpler statistical tools fail to do so, and their wide range of applications, the last 25 years have seen many advances in the joint modeling field. Even though interest and developments in joint models have been widespread, information about them has been equally scattered in articles, presenting recent advances in the field, and in book chapters in a few texts dedicated either to longitudinal or survival data analysis. However, no single monograph or text dedicated to this type of models seems to be available. The purpose in writing this book, therefore, is to provide an overview of the theory and application of joint models for longitudinal and survival data. In the literature two main frameworks have been proposed, namely the random effects joint model that uses latent variables to capture the associations between the two outcomes (Tsiatis and Davidian, 2004), and the marginal structural joint models based on G estimators (Robins et al., 1999, 2000). In this book we focus in the former. Both subfields of joint modeling, i.e., handling of endogenous time-varying covariates and nonrandom dropout, are equally covered and presented in real datasets''
