Introducing Survival and Event History Analysis

Md Kamrul Islam
2014 Canadian Studies in Population  
In this book, Melinda Mills aims to introduce survival and event history analysis by covering a wide range of topics to non-specialists and specialists. What makes the book special is the wide range of topics coveredranging from non-parametric to parametric methods. The book is most instructive for illuminating the use of popular statistical packages, such as R and Stata, for survival and event history analysis. Whereas most books on statistical analysis focus on either theoretical aspects or
more » ... atistical features, this book intends to bridge the two. Considering the increasing interest in survival and event history analysis in the social sciences, and the availability of sophisticated statistical packages such as Stata and R, the timing of the book is just right. The book is divided into 11 chapters. Chapter 1 commences with a brief introduction of the survival and event history analysis. Mills clearly defines the key concepts and terminologies, and explains the importance of studying survival and event history analysis. Then, in Chapter 2, the author provides excellent guidance on how to use the statistical package R-one of the most widely used data analysis software in recent years. In this chapter, the author focuses on the advantages and disadvantages of using different statistical programs for survival and event history analysis, shows how to download instructions of the base program R, explains three different approaches of using R, and provides empirical examples of using R to analyse survival data. The chapter is very well written and easy to follow. The first difficulty that researchers in general and lay researchers in particular face is preparing datasets for analysis, a topic that has received limited attention in most books. Mills addresses the components of rearranging and preparing datasets for survival and event history analysis in a highly readable manner in Chapter 3. Three types of data are mostly used in survival and event history analysis: single-episode data, multi-episode data, and subject-period (discrete-time) data. The particular focus on converting single-episode data into multiple-episode data and transforming date variables into numeric variables is very useful for new researchers. The starting point of survival and event history analysis after rearranging and preparing the dataset is to obtain the Kaplan-Meir survival estimates (KM) and plot them to get a first idea about the effect of exposure variable of interest on the outcome variable. Chapter 4 clearly illustrates the computing of KM survival estimates, plotting the KM survival curves, and comparing two KM survival curves. This chapter would guide researchers in selecting the most appropriate test in examining differences between two groups. The limitation of the KM survival estimates is that they fail to take other covariates into account, which can be done through the application of Cox proportional hazards (PH) regression. Chapter 5 explains why the Cox PH model is so popular as compared to other parametric models. The procedures for estimating and interpreting the Cox model, with both fixed and time-varying covariates, are clearly outlined. Chapter 6 is devoted to specification of five parametric models: exponential, piecewise constant exponential, Weibull, log-logistic, and log-normal models. The chapter also contains a specific focus on the relationship between probability density function, hazard function, and survival function. Mills explains the estimation and interpretation of each of the parametric models, along with the relative advantage and disadvantage of using
doi:10.25336/p6v89x fatcat:aqxgvhta3nebfe2lr52rrxuep4