Customer Reviews

Applied Survival Analysis: Regression Modeling of Time to Event Data

5 star:		(6)
4 star:		(2)
3 star:		(0)
2 star:		(0)
1 star:		(0)

 

 

The authors provide a really nice, non-technical survey of the landscape for Cox Proportional Hazards models. A nice aspect of their treatment is the care they take to reference all highly technical texts and journal articles. For example, if you'd like to find out more about goodness-of-fit tests for survival models, the authors provide ample references to the Counting Process Theory of Martingale Residuals.

The first chapter discusses the basic characteristics of survival data, including the notion of censoring (in all of its various forms). Examples of the principle types of censoring are included. The chapter also includes introductory material on the general survival model, including a nice description of the log likelihood function. Curiously, the rigorous definition of the hazard function has been omitted, probably to avoid intimidating readers who are not familiar with formal limits.

Chapter 2 continues to build up the general survival model and introduces the relationship between the survivor function and the cumulative hazard. Pointwise estimators for the survivor function are discussed, including the Kaplan-Meier estimator along with the various variance estimators. Test statistics for comparing two survival populations are introduced, including the Log-Rank and General Wilcoxon statistics. The reader is encouraged to read the counting process treatments of these statistics to see why they produced defensible hypothesis tests.

Chapter 3 is devoted to the Cox Model and Cox's partial likelihood function. Tests for significance of the coefficients are introduced, included the Wald test, log likelihood ratio test and the score test. These are used heavily in the later chapters as the basis of a model-building methodology.

Chapter 4 is a very short, but nicely written chapter explaining how to interpret the values of each regression coefficent. It also describes covariate-adjustment techniques for model diagnostics.

Chapter 5 is just a wonderful chapter which outlines classical model building techniques. This is a great chapter for anyone who has ever been thrown a ton of data (with a bushel of possible covariates) and asked to "fit a model to this stuff".
Readers who have done a lot of purposeful fitting of linear regression models won't find the basic techniques new, but use of survival specific residuals and selection criterion will probably be an eye-opener. The section on assessing the functional form for continuous covariates is also nicely written.
However, the section on Best Subsets Selection was a little too "cook-booky" for my taste.

Chapter 6 is another very nice chapter on goodness-of-fit. It discusses analysis of the various residuals and their use for analysis outliers, testing proportional hazards assumptions and overall Goodness-of-Fit.

Chapter 7 discusses the standard extensions of the Cox model, including stratification and time-varying covariates. Chapter 8 discusses parametric survival models, and is a good introduction to the SAS procedure LIFEREG. The generalization of the Cox model to recurring event data (also know as Aalen's multiplicative intensity model) can be found in Chapter 9.

My only complaint is that each chapter was designed to be read in one sitting. Individual ideas, topics and formulas can be buried in a seemingly unbroken chain of paragraphs. The lack of sub-sub section titles,etc, makes using the text as is somewhat cumbersome to use as a desk reference. I've gotten around this limitation by marking key concepts, etc., in the margin in order to give a "quick search" capability enhancement to the index.

Applied Survival Analysis is an excellent book for someone seeking a non-mathematicial explanation of survival analysis. The book covers the motivation behind the development of survival analysis, estimation of survival curves, the Cox proportionial hazards, and some parametric models. The book also covers the major methods used in variable selection, model building, and diagnostics. Someone with an undergraduate background in statistics and econometrics will understand the book. The book relies on text to discuss the methods and uses mathematical formulas only when absolutely necessary. Numerous examples are used to highlight what the text covers. The math that is used is easily understandable. This book is ideal for someone who needs to learn the tools of survival analysis but not how they were derived.

I enjoyed the authors' book on logistic regression analysis in 1989, and this book is just as good, or better, with many extremely practical suggestions on building regression models for survival data. Happily, the authors summarize, compare, and contrast several major texts on survival analysis which have appeared in the past 10 years. For example, they discuss different names used by different authors for score residuals. They present a helpful appendix on the counting process approach to survival analysis, which will make more advanced texts accessible to students; thus, anyone who wants to use survival analysis, at any level, should consult this book, even if he has already studied books by Miller, Lee, Collett, Fleming-Harington,Andersen, et al, etc. An unfortunate drawback to this book is that the first printing contains many careless errors, some of which may affect student learning: for example, the definition of a survival function is misstated. I recommend that you insist on the second or third printing when buying this book, and you will be quite satisfied.

This book provides a good, clear, concise explanation of Cox's proportional hazards models. For someone seeking a non-mathematical description this is a great guide. The original datasets from the text examples can even be downloaded and you can go through the same process yourself. Because of some mistakes in the text, I would recomend looking at other sources as well.

Hosmer and Lemeshow have given us a clear, nontechnical introduction to using survival models. The book strikes a good balance between covering the basics and addressing the most recent, state-of-the-art techniques, including repeated events, frailty models, and others. They also do a good job of addressing practical issues, including estimation details and available software. While most of the examples are drawn from medicine and biostatistics, this book could also serve as a useful starting point for social and behavioral scientists interesting in learning the fundamentals of these models, as well as a useful reference for applied researchers.

Would definitely buy this again. It was a good read and its principles can be applied to many different areas. It is a good resource when analyzing failure data.

It arrived fast, which I can study it in time. The deal can take place even I do not have usa account in my mastercard, so It is convenient to me.

The book arrived in a timely manner and was as expected based upon the available preview. Nice improvements were made to this edition.