Contributed by Paul Allison, Professor of Sociology, University of Pennsylvania

I'm happy to announce that the second edition of my book Survival Analysis Using SAS: A Practical Guide has just been published by the SAS Institute.

When the first edition was published in 1995, my goal was to provide an accessible, data-based introduction to methods of survival analysis, one that focused on methods available in SAS and that also used SAS for the examples. The success of that book confirmed my belief that statistical methods are most effectively taught by showing researchers how to implement them with familiar software using real data.

Of course, the downside of a software-based statistics text is that the software often changes more rapidly than the statistical methodology. In the 15 years that the book has been in print, there have been so many changes in the features and syntax of SAS procedures for survival analysis that a new edition was long overdue. So it’s a great relief that I no longer have to warn potential readers about out-of-date SAS code.

Although the basic structure and content of the book remain the same, there are numerous small changes and several large changes. One global change is that all the figures use ODS graphics. Here are some other major changes and additions:

Chapter 3, “Estimating and Comparing Survival Curves with PROC LIFETEST.” This chapter documents some major enhancements to the STRATA statement, which now offers several alternative tests for comparing survivor functions. It also allows for pairwise comparisons and for adjustment of p-values for multiple comparisons. In the first edition, I demonstrated the use of a macro called SMOOTH, which I had written to produce smoothed graphs of hazard function. That macro is no longer necessary, however, because the PLOTS option (combined with ODS graphics) can now produce smoothed hazard functions using a variety of methods.

Chapter 4, “Estimating Parametric Regression Models with PROC LIFEREG.” This chapter now includes a section on the PROBPLOT command, which offers graphical methods to evaluate the fit of each model. The last section introduces the new BAYES statement which, as the name suggests, makes it possible to do a Bayesian analysis of any of the parametric models using MCMC methods.

Chapter 5, “Estimating Cox Regression Models with PROC LIFEREG.” The big change here is the use of the counting process syntax as an alternative method for handling time-dependent covariates. When I wrote the first edition, the counting process syntax had just been introduced, and I did not fully appreciate its usefulness for handling predictor variables that vary over time. Another new topic is the use of the ASSESS statement to evaluate the proportional hazards assumption. Also new is a section on customized hazard ratios, which are especially useful for interpreting interactions. Finally, there is a section that describes the BAYES statement for estimating Cox models and piecewise exponential models.

Chapter 6, “Competing Risks.” This chapter now contains a section on cumulative incidence functions, a popular alternative approach to competing risks.

Chapter 7, “Analysis of Tied or Discrete Data with the LOGISTIC procedure.” The first edition also used the PROBIT and GENMOD procedures to do discrete time analysis. But PROC LOGISTIC has been enhanced to the point where the other procedures are no longer needed for this application

Chapter 8, “Heterogeneity, Repeated Events, and Other Topics.” For repeated events and other kinds of clustered data, the WLW macro that I described in the first edition has been superseded by the built-in option COVSANDWICH. In this chapter I also describe the use of the new GLIMMIX procedure to estimate random-effects models for discrete time data.