The statistical analysis of recurrent events. (Q2384380): Difference between revisions

This book deals with processes generating multiple events over time. Conventional survival analysis tools will be rarely adequate for such recurrent events data, so that more appropriate models and statistical methods are necessary. The book comprises eight chapters, four appendices and a useful notational glossary. Bibliographic notes are provided for all chapters, and problems and supplements are provided for all but chapter 1. Solutions to the problems are not included in the book. Chapter 1 outlines the scope of the book by providing vivid real-life examples and introducing key concepts in brief. Basic types of recurrent events data are based on either counts or gap times, which results in Poisson and renewal processes, respectively. Chapter 2 introduces both processes and touches on possible extensions for them. Chapter 3 addresses count-based Poisson processes in great detail. Parametric, semi- and nonparametric, as well as random effects Poisson models are studied among others. Extensions which are based on marginal rate or mean functions are also contained in Chapter 3. Gap time based renewal processes, extensions for them, and specific estimation issues are topics of Chapter 4. Chapter 5 is devoted to general intensity-based models where the intensity function depends on previous event histories which, among others, could be previous event counts or gap times. The log-linear intensity and the trend-renewal model are two parametric exponents of this class of models. Semiparametric approaches rest upon Markov or modulated renewal processes. Nelson-Aalen and Aalen-Johansen estimates as well as stratified Andersen-Gill and Cox models come into play. Chapter 6 deals with multivariate recurrent events data. Multivariate extensions of specific models are mostly required if the events in question are unilaterally or reciprocally dependent. Of particular interest are recurrent events with a terminal event and alternating two-state processes. Chapter 7 addresses incomplete and selective data structures, in particular intermittent, dependently censored and event-dependently selected observations. Chapter 8 contains a conglomeration of practical issues, of which prediction and randomized trials are the most important ones. Appendix A contains a brief summary of maximum likelihood methods and estimating functions. Appendix B touches on computational methods. In Appendix C, code for S-PLUS or R software is given for selected examples, and Appendix D contains some data sets. There is an accompanying homepage for the book at {http://www.stats.uwaterloo.ca/cook-lawless/} which provides the table of contents, data sets, statistical software code, a list of known errata and typos, and links to additional resources. Although the book appeared in the Springer series ``Statistics for Biology and Health'', it is directed to a much broader target readership, like social scientists, economists and industrial statisticians as well. Elementary knowledge of survival analysis methods and counting process notations are essential for the reader. Nevertheless, the book has a strong applied focus. Many examples are used to illustrate and discuss the models and statistical methods in great detail. Techniques for estimation, testing and model checking are lucidly described so that the book could provide a solid basis for a graduate course. To sum up, the book has the potential to become a classic in the field of recurrent events analysis.