The statistical analysis of interval-censored failure time data. (Q2494658)

This book is a survey of statistical analysis tools for interval-censored data which occur rather frequently in various empirical sciences, e.g., in demography, epidemiology, finance, medicine or sociology. The data is interval-censored when one does not know the exact time point of the occurrences of the event of interest (which is often the survival of the subjects under investigation) but one only knows a time window of its occurrences. Exact or right-censored failure or survival times are special cases of interval censored failure times. There exist multivarious literature and methods for right-censored survival data which, however, can not be applied nor even easily extended for the analysis of interval-censored data. Chapter 1 introduces the different types of censored data (e.g., right-censored, left-censored, case I interval-censored, case II interval-censored, doubly censored, panel count data) and gives an overview of the basic concepts. Furthermore, regression models for the analysis of failure time data are considered, in particular the proportional hazards model, proportional odds model, additive hazards model, accelerated failure time model, linear transformation model and the discrete regression model. The different types of data are presented as the lead examples of the book illustrating interval-censored data in various case studies which are referred to in the following. In chapter 2 some parametric models (exponential, Weibull, log-normal, log-logistic) are considered and single point as well as multiple imputation methods are discussed to handle missing data. The main part of the book are chapters 3 to 10 dealing with nonparametric and semiparametric statistical approaches for interval-censored data which are in the focus of the book. Chapter 3 considers statistical procedures for nonparametric estimation of survival and hazard functions. Chapter 4 presents nonparametric comparison approaches of survival functions by means of rank-based (generalized log-rank test) as well as survival-based procedures for discrete and continuous survival times. Next, regression analysis of case I interval-censored and case II interval-censored data are given in chapter 5 and chapter 6, respectively. Hereby the earlier introduced regression approaches are considered, e.g., the proportional hazards and odds model or the additive hazards model for case I interval-censoring. For case II interval censoring the author addresses proportional hazards and odds models as well as accelerated failure time and logistic models. Chapter 7 presents nonparametric and semiparametric approaches for the analysis of possibly correlated bivariate interval-censored failure times. A main focus in this context is to measure the association of these two variables. The author discusses, e.g., parametric copula models and nonparametric maximum likelihood estimation of the joint distribution. Chapter 8 deals with doubly censored failure times data which characterize the duration between two related events which both can be right- or interval-censored. This type of data arises, e.g., as one is interested in the time between infection and the onset of a certain disease if the time of infection cannot be observed exactly. Nonparamatric approaches are considered to estimate distribution functions and to compare several survival functions. Several semiparametric regression methods are also discussed (discrete proportional, continuous proportional and additive hazards model). The analysis of panel count data can finally be found in chapter 10. This type of data is observed when an event or failure occurs at discrete time points resulting in numbers of occurrence of events between observation times. To analyze panel count data the occurrence of recurrent events is characterized by counting processes and their mean function. Methods of nonparametric inferences (estimation and comparison) are considered as well as regression analysis for estimation of mean functions. Chapter 10 contains brief discussions of several special issues including regression diagnostics, regression analysis with interval-censored covariates, Bayesian inference approaches and informative interval-censoring. Each chapter of the book contains several examples and it is shown how to apply the presented methods to real data sets. Often the numerical values of the interesting observations are given in tables encouraging the reader to recalculate and reproduce the results. Additionally, five complete sets of data discussed in the book are provided in the appendix of the book. Each chapter ends with bibliographical notes and numerous references. Algorithms for fitting the presented models should facilitate the computational implementation of the methods. A short review of available statistical software can be found in the last section of chapter 10. This volume is adequate for self-study for researchers and students who want to handle and analyze interval-censored data.