Abstract
Intensive longitudinal data are increasingly widely collected to permit researchers to investigate important research questions on intraindividual change, variability, and relations and to further understand interindividual differences in these change processes (Baltes & Nesselroade, 1979; Molenaar, 2004). Bolger and Laurenceau provide a timely, easy to follow, practically oriented, and useful introduction to frequently used models and important issues in intensive longitudinal research. The coverage of the issues and models is thorough and clear. More specifically with respect to modeling, univariate, bivariate, and trivariate models are introduced with various types of data from independent individuals or nonindependent dyadic members (e.g., wives and husbands). In addition to introducing the necessary statistical models for analyzing intensive longitudinal data, Bolger and Laurenceau discuss fundamental measurement and design issues for planning and designing intensive longitudinal studies. The book is a valuable and useful resource for methodologists, applied researchers, and students of intensive longitudinal research alike. It is especially useful due to its provision of (1) concrete examples of how to implement each data analytic strategy and interpret the outputs and results using example data sets and common statistical software and (2) clear and well-written example write-ups. Moreover, the one-sentence description of each recommended reading at the end of each chapter is a practical way for busy readers to select the most relevant follow-up literature to read. The following provides a review of the content of each chapter in the book.
In the opening chapter, Bolger and Laurenceau introduce readers to the rationale and organizational framework of the book. They provide a summary of benefits of intensive longitudinal methods, a definition of intensive longitudinal designs, and an overview of applications of intensive longitudinal methods. Bolger and Laurenceau describe that the purpose of the book “is to provide basic guidance on these issues so that readers can plan, carry out, analyze, and publish their own intensive longitudinal studies” (p. 7). The authors have successfully achieved these goals by discussing various aspects of intensive longitudinal methods in Chapters 2 through 10.
In Chapter 2, “Types of Intensive Longitudinal Designs,” Bolger and Laurenceau discuss the overall strengths of intensive longitudinal designs. They detail the differences among four design subcategories by reviewing major characteristics, providing examples of studies, and discussing relevant limitations of each type. Bolger and Laurenceau recommend considering various factors such as continuous versus discrete events, participation burden, frequency of assessments, rare versus frequent events, and resources when selecting the most appropriate design for a specific study. This chapter is valuable and useful for the early stages of planning an intensive longitudinal study.
Five basic guidelines of intensive longitudinal data analysis are described and reviewed in Chapter 3, entitled “Fundamentals of Intensive Longitudinal Data.” Enumerating these five guidelines early in the book allows readers to appreciate their importance: Ignoring any one of them would lead to either less interpretable or even misleading results. They (1) distinguish the between-subject and within-subject levels of analysis, (2) allow for between-subject random effects, (3) consider the influence of time, (4) specify the appropriate number of independent units, and (5) choose interpretable zero points for within-subject predictors. Bolger and Laurenceau use a hypothetical example data set and vivid graphs to explain the five guidelines. The five guidelines as well as their importance are well summarized. In the following chapters, the authors consistently follow the five guidelines to introduce and explain the models and methods.
In Chapter 4 entitled “Modeling the Time Course of Continuous Outcomes,” Bolger and Laurenceau use an example of a two-group intervention study with weekly diary assessments on intimacy to answer two basic kinds of questions regarding the time course: intraindividual change and group differences in intraindividual change. An application of linear growth curve analysis is provided with (1) clear explanations of fixed effects and random effects in the model, (2) readily applicable example syntax, (3) easy to follow interpretations of the outputs from software, and (4) a helpful example write-up. Instead of assuming that Level 1 residuals are independent over time, Bolger and Laurenceau override the ordinary default and use a first-order autoregressive error covariance structure to model the Level 1 residuals, which is more general and flexible. This covariance structure is also applied in the following chapters when applicable. The material is useful for substantive researchers to understand multilevel analysis at its most basic level, before adding in additional components. Additionally, the write-up example provides a convenient template for students and researchers to use for describing their own growth curve modeling results.
Chapter 5, “Modeling the Within-Subject Causal Process,” covers a brief overview of causal inference as well as essentials and procedures of studying the relations between a time-varying predictor (daily relationship conflicts in the chapter) and a time-varying outcome variable (daily intimacy in the chapter). Bolger and Laurenceau review two guidelines discussed in Chapter 3 (i.e., distinguishing the between-subject and within-subject levels of analysis and considering the influence of time) and apply them in this chapter. Regarding the first guideline, they emphasize the importance of person-mean centering the time-varying predictor for disaggregating between- and within-subject relations using the hypothetical example (e.g., Curran & Bauer, 2011). Regarding the second guideline, Bolger and Laurenceau advise that “elapsed time should always be included in the model” (p. 71). Wang and Maxwell (2015), however, recommend that time may or may not need to be included, depending on the longitudinal study design and the research question of interest. Nevertheless, the provided example code and the example write-up are clear and useful for readers to analyze and summarize similar studies. In addition, Bolger and Laurenceau briefly discuss how to handle missing data and unequal time intervals, two frequently encountered problems in intensive longitudinal research. In future editions of the book, a potentially beneficial inclusion would be an entire chapter dedicated to more thorough coverage of missing data and other commonly occurring hurdles found in intensive longitudinal methods.
In Chapter 6, “Modeling Categorical Outcomes,” Bolger and Laurenceau provide step-by-step procedures of modeling the relationship of a time-varying predictor (female partner’s morning anger in the chapter) and a time-varying binary outcome (male partner’s report of interpersonal conflict later that day, yes or no, in the chapter). Due to the nature of the categorical data, fundamental concepts such as the relationship between the mean and variance of a binary variable, underdispersion, and overdispersion are discussed. The disaggregation of within- and between-subject relations are also appropriately considered in the generalized multilevel models. The SAS, SPSS, and Mplus code should be useful for applied research with categorical outcomes to apply models alike to their own data.
Chapter 7 is on “Psychometrics of Intensive Longitudinal Measures of Emotional States.” The inclusion of a psychometric chapter is appreciated, as the importance of reliably estimating change is often underemphasized in other texts on intensive longitudinal approaches. Bolger and Laurenceau conceptually discuss measures of reliability from classical test theory, generalizability theory, and multilevel confirmatory factor analysis perspectives for intensive longitudinal research. Bolger and Laurenceau appropriately utilize the latter two frameworks to illustrate the concepts and calculation of the reliability of within-subject changes in daily positive mood assessed with multiple items from a daily diary study. The model equations and reliability formulas are introduced and explained logically so that faculty and students can easily grasp the essence of the concepts. The syntax is useful for researchers to apply the methods to calculate reliabilities for their own studies. Additionally, the recommended readings are excellent supplementary resources for interested readers who desire to learn more about the topic.
In Chapter 8, “Design and Analysis of Intensive Longitudinal Studies of Distinguishable Dyads,” Bolger and Laurenceau begin with a discussion of important and compelling topics of intensive longitudinal dyadic research. They emphasize the necessity of analyzing dyad members simultaneously and clarify the use of two-level instead of three-level models for typical longitudinal dyadic data analysis. An example with clear and useful model equations, code, result interpretations, and sample write-up is provided. In the example, they study how changes in daily work stressors of one partner of a couple are linked to changes in relationship dissatisfaction of the other partner over time. They also provide detailed and convincing explanations of why the correlations between partner’s residuals at the within-couple level and correlations between partner random effects at the between-couple level are allowed to be nonzero and need to be freely estimated.
Chapter 9, “Within-Subject Mediation Analysis,” discusses mediation analysis at the within-subject level. As an example, Bolger and Laurenceau use a 1-1-1 multilevel mediation model to study whether and how Level 1 work dissatisfaction mediates the relationship between Level 1 work stressors and Level 1 relationship satisfaction intraindividually. They continue to appropriately use person-mean centering to disaggregate within- and between-person effects in the multilevel mediation analysis (e.g., Preacher, Zyphur, & Zhang, 2010). Both the SAS and Mplus code and the results are clearly explained and generalizable to readers’ own research.
The final chapter (Chapter 10), “Statistical Power for Intensive Longitudinal Designs,” covers the often overlooked sample size determination issue. Bolger and Laurenceau first generally review the power issue in multilevel models. The sampling variance formula (p. 200) is helpful and clearly shows that the power in multilevel analysis is influenced by both within-subject and between-subject variances. The authors’ discussion of how the number of subjects and number of time points play roles in influencing the two kinds of variances is intriguing. Then, Bolger and Laurenceau use the Monte Carlo power simulation function in Mplus to illustrate how to conduct power analysis for all the models discussed in the previous chapters. The provided code and explanations of the results are extremely applicable to researchers conducting sample size determination for planning their own studies.
Bolger and Laurenceau present an excellent and extensive introduction for researchers and students who either have collected or plan to collect intensive longitudinal data. The coverage of the topics is conceptually comprehensive yet practical for researchers to plan a future intensive longitudinal study and analyze intensive longitudinal data. In future editions of the book, however, we suggest that the authors consider adding chapters on other models for intensive longitudinal analysis such as dynamic system modeling (e.g., Boker & Nesselroade, 2002; Chow, Ferrer, & Nesselroade, 2007), state-space modeling (e.g., Kitagawa, 1996; Song & Ferrer, 2009), and continuous time modeling (Oud & Jansen, 2000) to make the book even more comprehensive for more advanced readers.
Two notable strengths of the book pertain to its simplicity of presentation and attention to pragmatic considerations. Specifically, the SPSS, SAS, and Mplus example code and example data sets for various models and methods provided throughout the book are very beneficial for both learning and teaching. In addition, Bolger and Laurenceau use interesting examples in social sciences to aid readers in understanding the models and methods. A unique feature of this book is that the examples in different chapters are conceptually related to each other, illustrating how small individual pieces of intensive longitudinal research can form a rich bigger picture when considered together. For example, Chapter 4 studies change in intimacy over time, Chapter 5 examines the longitudinal relations between conflict and intimacy, whereas Chapter 6 predicts the occurrence of conflict from anger. Furthermore, Bolger and Laurenceau study the bivariate relationship between daily work stressors and relationship dissatisfaction in Chapter 8 and then extend the bivariate relationship to trivariate by seamlessly including work dissatisfaction in Chapter 9 as a mediator to explain the previously defined bivariate relationship. In addition, the models in Chapter 10 for power analysis are based on all of those discussed in the previous chapters. Therefore, the chapters follow a natural progression to facilitate learning and the book presents a coherent picture of intensive longitudinal methods.
The authors supplement the book with a user-friendly website that contains not only the syntax and data files presented in the book but also sample code for programs not presented in the text itself. Specifically, the website contains well-organized sample syntax for SPSS, SAS, Mplus, Stata, HLM, and R as well as data files saved in formats that correspond to the syntax format. The authors clearly understand that readers of this book may need to use a variety of different programs to analyze intensive longitudinal data. By including an array of syntax languages, Bolger and Laurenceau increase the practical utility of the book even further.
Bolger and Laurenceau have provided a book that will prove extremely welcoming for students learning intensive longitudinal methods and for researchers conducting intensive longitudinal studies. We would recommend this book as a textbook for an intensive longitudinal methods course and a valuable supplement for an applied longitudinal analysis course or an advanced longitudinal methods course. Instructors and students will enjoy the strong focus on application with the step-by-step examples and practical guidance. Readers with background knowledge of regression and analysis of variance will be able to teach themselves the basics of the models using the helpful example code and data sets and scientifically record their own results following the template of the example write-ups. Overall, we believe this book should be considered an essential reference for any researcher engaging in intensive longitudinal research.