models for intensive longitudinal data

2
This article was downloaded by: [Carnegie Mellon University] On: 09 November 2014, At: 03:31 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK Journal of the American Statistical Association Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/uasa20 Models for Intensive Longitudinal Data Daniel B Hall a a The University of Georgia Published online: 01 Jan 2012. To cite this article: Daniel B Hall (2007) Models for Intensive Longitudinal Data, Journal of the American Statistical Association, 102:480, 1473-1473, DOI: 10.1198/jasa.2007.s228 To link to this article: http://dx.doi.org/10.1198/jasa.2007.s228 PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions

Upload: daniel-b

Post on 13-Mar-2017

213 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Models for Intensive Longitudinal Data

This article was downloaded by: [Carnegie Mellon University]On: 09 November 2014, At: 03:31Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: MortimerHouse, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of the American Statistical AssociationPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/uasa20

Models for Intensive Longitudinal DataDaniel B Hallaa The University of GeorgiaPublished online: 01 Jan 2012.

To cite this article: Daniel B Hall (2007) Models for Intensive Longitudinal Data, Journal of the American StatisticalAssociation, 102:480, 1473-1473, DOI: 10.1198/jasa.2007.s228

To link to this article: http://dx.doi.org/10.1198/jasa.2007.s228

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose ofthe Content. Any opinions and views expressed in this publication are the opinions and views of the authors,and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be reliedupon and should be independently verified with primary sources of information. Taylor and Francis shallnot be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and otherliabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to orarising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Page 2: Models for Intensive Longitudinal Data

Book Reviews 1473

Models for Intensive Longitudinal Data.

Theodore A. WALLS and Joseph L. SCHAFER (eds.). New York: OxfordUniversity Press, 2006. ISBN 0-19-517344-9. xxii + 288 pp. $65.00.

Modern information technology now allows the easy collection and storageof massive amounts of data. This development presents tremendous opportu-nities and challenges for scientists and statisticians. Traditionally, longitudi-nal study designs have most often used relatively short series of measurementsthrough time (e.g., ≤10 measurement occasions), and statistical methodologyfor longitudinal data has concentrated on this case. However, recent technolog-ical innovations have led to the increasing popularity of studies that generatewhat Walls and Schafer term “intensive longitudinal data” (ILD). Such data,stemming from many (typically frequent) measurement occasions through timeon a sample of subjects, arise in various disciplines. However, it is their growingprevalence in psychology and related areas of social science that has providedthe primary motivation for the publication of Models for Intensive LongitudinalData.

This edited volume contains a collection of chapters that describe and illus-trate statistical models and methodology appropriate for ILD. The contributingauthors are mostly social scientists and statisticians, with many of the chapterswritten collaboratively across these disciplines. The emphasis is on explainingvarious classes of models: their definition, conceptualization, interpretation, re-lationship to other model classes, and usefulness for addressing questions ofscientific interest. Technical, theoretical, and other aspects of limited interest topractitioners are kept to a minimum and, in some cases, relegated to appendixes.Nearly all of the chapters include applications that are pursued further than istypical in statistical publications, with novel substantive results presented inmany cases. Topics represented in the applications include smoking behavior,patterns of alcohol use, relationship of feelings of control and choice over dailyactivities, marital intimacy, daily patterns of personality characteristics, neuralmechanisms of attentional control, and traffic intensity.

After a nicely written introduction that motivates, identifies themes in theanalysis of ILD, and summarizes the chapters, the editors and their collabora-tors contribute two chapters that introduce and summarize the main ideas oftwo traditional methodologies for longitudinal data: the linear mixed model(also called multilevel model or hierarchical linear model) and generalized es-timating equations. This material sets the stage for subsequent chapters thatdiscuss functional multilevel models, item response models and their relation-ship to mixed models, semiparametric varying-coefficient linear mixed models,multilevel autoregressive models, state-space models, dynamical systems mod-els based on differential equations, and point process models. A final chapterhighlights some emerging technologies for the collection of ILD and discussesthe associated statistical challenges.

All of the chapters are written at an accessible level. Collectively, they do anice job of bridging the gap between the social sciences and the field of statis-tics, clarifying the relationship between methodology and terminology used inthe two areas while introducing a broad array of models and methods for ILDanalysis. Although this volume will be of greatest value to quantitatively ori-ented researchers in psychology and other areas of social science, I also recom-mend it to statisticians and anyone else interested in the collection and analysisof ILD.

Daniel B. HALL

The University of Georgia

The Statistical Analysis of Interval-Censored Failure Time Data.

Jianguo SUN. New York: Springer, 2006. ISBN 0-387-32905-6. xv +302 pp. $79.95.

The analysis of failure time data is one of the most active research areas instatistics. The time to the event of interest is the main concern in this type ofanalysis. However, quite often, unexpected or expected termination of follow-up (right-censoring) occurs. Thus an extensive amount of work has focused onright-censored failure time data. Interval-censored failure time data, where thefailure time of interest often is not observed but is known only to fall withinsome interval, occurs commonly in clinical trials and medical studies. In manysituations, these intervals are irregularly spaced and differ from patient to pa-tient. Interval censoring, a more general case of right-censoring, commonly

arises when subjects are not under continuous observation yet undergo peri-odic examination instead. Consequently, the failure time cannot be observedexactly but is only known to lie between two examinations.

Well-known textbooks concerned with failure time data analysis includethose by Cox and Oakes (1984), Kalbfleisch and Prentice (2002), and Kleinand Moeschberger (2003). Because the analysis of interval-censored data is arelatively new research area, most textbooks devote only one or two chapters tothe subject. This is the first book that provides a comprehensive summary of im-portant topics in the field; therefore, there is no doubt that this book representsan important contribution to the literature.

The book is divided into 10 chapters that can be divided into 4 distinct parts.Chapter 1 begins with motivating examples and a gentle introduction to generalcensoring and interval-censored failure time data. Subsequent material illus-trates types of interval-censored data and a collection of possible hazard modelsthat can be applied.

The second part, comprising Chapters 2–6, considers methodologies fordealing with incompleteness of univariate interval-censored data. Chapter 2 dis-cusses likelihood-based parametric models and imputation approaches. Chap-ter 3 is devoted to nonparametric maximum likelihood estimation for survivaland hazard functions. Chapter 4 covers the different methods of comparisonfor survival functions. Finally, Chapters 5 and 6 discuss regression analysis ofcase I (current status) and case II (general, interval-censored) data. Current sta-tus data occur when each study subject is observed only once, and informationis available as to whether the event has occurred no later than the observationtime.

Chapters 7–9 cover additional features of interval-censored data. Chapter 7introduces several algorithms for analyzing bivariate interval-censored data.Chapter 8 is concerned with doubly censored failure time data, where two re-lated failure events occur, one followed by the other. Chapter 9 discusses panelcount data, which are interval-censored event history data. This type of data isa generalization of current status data.

The last part of the book, Chapter 10, features such specialized topics as re-gression diagnostics for parametric models and certain semiparametric models,interval-censored covariates, Bayesian analysis, and informative interval cen-soring.

The author states that this book will be particularly useful for graduate stu-dents in statistics or biostatistics who have a basic knowledge of probability andstatistics. However, I think that a one-semester course in failure time analysiswould serve as a helpful prerequisite for understanding the covered material.The text also should provide a valuable reference for researchers who seek anoverview of the main topics pertaining to interval-censored data, as well asthe methods available for analyzing such data. A basic understanding of im-putation, the expectation-maximization (EM) algorithm, kernel estimation, andother related topics, would facilitate a better comprehension of some of thefeatured topics.

Due to the complex nature of interval-censored data, it could be a difficultchallenge to present the subject in an accessible fashion, but the author ad-mirably meets this challenge. His goal seems to be to present a balance betweentheory and applications. Many of the book’s main results are not proven, partic-ularly those requiring long or intricate derivations, but relevant citations to theliterature are consistently provided. (The bibliography contains more than 400references.) The treatment of some of the theoretical material is rather terse,especially the coverage pertaining to asymptotic results.

The book is software-independent, but all of the chapters except the last one(which covers specialized topics) present examples with real data. In general,each topic is introduced by first motivating the basic ideas and methods andthen applying these methods to examples. The book does not provide computercode but does mention that most of the computational work, for examples, isconducted using S–PLUS and R. In addition, some references are provided forcomputation.

In summary, this book is good news for those interested in an overview ofinterval-censored data. It emphasizes the unifying role of statistical models andmethodologies for interval-censored data, avoiding unnecessary mathematicaldetails yet still providing extensive references. This book offers a well-writtenand comprehensive summary. I enjoyed reading it and feel that it provides anexcellent guide to this important area. It will definitely be kept on my shelf forreference.

Do-Hwan PARK

University of Nevada–Reno

Dow

nloa

ded

by [

Car

negi

e M

ello

n U

nive

rsity

] at

03:

31 0

9 N

ovem

ber

2014