The investigation of change over time is key to empirical research in many disciplines. What factors influence growth in African American students’ academic achievement over the course of college? When and why do families leave Early Head Start? Such research questions allow for the generation of useable knowledge through empirical research. In their latest book, Applied Longitudinal Data Analysis, Judy Singer and John Willett make the case that such research questions are best answered through the collection and analysis of longitudinal data, while also providing a clear and concise tutorial on the use of two methods therein: individual growth modeling and survival analysis.

Although appropriate statistical methods for the analysis of longitudinal data were first developed during the 1980s, they have yet to be widely applied in educational and psychological research. This is unfortunately true, even when researchers make the effort to collect data at multiple time points. As Singer and Willett note, “In a review of over 50 longitudinal studies published in American Psychological Association journals in 1999, for example, we found that only four used individual growth modeling (even though many wanted to study change in a continuous outcome) and only one used survival analysis (even though many were interested in event occurrence)” (p. vii). By not making use of these methods in their work, researchers limit the kinds of effects that they are able to detect statistically, as well as the kinds of quantitative research questions they can ask.

It is likely that inattention to appropriate statistical methods for longitudinal data analysis in most popular applied statistical textbooks and the highly technical nature of the books that are available on these topics have contributed to their scant use in social science research. Singer and Willett tackle this problem head-on in two ways. First, they explain individual growth modeling and survival analysis step-by-step using real data. Such an approach makes these complex longitudinal analysis techniques understandable. Second, they explicitly and specifically target their tutorial for “our professional colleagues (and their students) who are comfortable with traditional statistical methods but who have yet to fully exploit these longitudinal approaches” (p. viii). In this way, Singer and Willett enter into “a structured conversation among colleagues” (p. viii) about the application of these methods instead of a complex technical discussion.

Singer and Willett divide their book into two main sections. The first section covers individual growth modeling (a special case of multilevel modeling). This methodology allows for the analysis of a continuous outcome, like heart rate or SAT score, and can include time as a predictor. Using this technique, researchers can ask and answer questions about how a particular variable changes over time and what factors are associated with that change. The second section covers survival analysis (otherwise known as event history analysis and hazard modeling in the literature). This methodology allows for the analysis of the occurrence of a specific event or dichotomous outcome, like drug abuse relapse or dropping out of an intervention program. In contrast to individual growth modeling, survival analysis allows the researcher to examine time as an outcome. For example, using this technique, one could investigate what factors are associated with whether and when divorce occurs; the occurrence of the event, divorce, in the context of time is the outcome.

Although the book is divided into two sections, it is important to note that Singer and Willett discuss growth modeling and survival analysis within a single framework to encourage the use of both methods by researchers in answering different longitudinal questions within the same study. Key ideas in data analysis are emphasized in both sections of the book, including the identification of appropriate research questions, the selection of sound methods of estimation, as well as the precise interpretation and presentation of results. To make these key ideas salient, the authors use real data donated by researchers across many disciplines (i.e., psychology, education, public health, and sociology) to demonstrate each method and to illustrate the interpretation of their findings in some detail.

In sum, Singer and Willett make complex statistical techniques understandable and useable through the use of “real data” examples, by emphasizing key ideas, and by applying a conversational tone. Applied Longitudinal Data Analysis has the potential to encourage researchers to ask longitudinal questions and then employ appropriate statistical techniques to begin answering them.

G.R.-S.