Design and Analysis of Time Series Experiments - Tapa blanda

Críticas

"This is a wonderful book for anyone especially interested in interrupted time-series design and analysis. The presentation is very clear, the main design and analysis options are comprehensive, and the advice that careful reading will elicit is invariably wise. I will recommend it to anyone in the social and behavioral sciences who wants a comprehensive and accessible introduction to what interrupted time series can achieve in the hands of those willing to trust this book."
--Thomas D. Cook, Joan and Serepta Harrison Emeritus Professor of Ethics and Justice and Professor Emeritus of Sociology, Psychology, Education, and Social Policy, Northwestern University

"McCleary, McDowall, and Bartos have advanced a field of research methodology that has lain fallow for more than three decades; and in the process, they challenge staid thinking in the area of verification of causal claims in disciplines as diverse as social science, economics, and medicine."
--Gene V. Glass, National Education Policy Center, University of Colorado

"McCleary, McDowall, and Bartos have truly extended our knowledge of time series methodology. And of special note, time series methodology is foundational for single-case intervention research and the causal inference associated with this unique methodology. McCleary and his fellow scientists have offered single-case researchers new and important insights into experimental and quasi-experimental causal inference. Most importantly, single-case design researchers will find new conceptual features of the time series validities that will further promote the credibility of this methodology in the social and behavioral sciences. I strongly recommend this important resource to my single-case research colleagues."
--Thomas R. Kratochwill, Sears-Bascom Professor of School Psychology, University of Wisconsin-Madison

"Time series data are ubiquitous nowadays. The innovations of this book relative to other time series texts are its integration of contemporary counterfactual approaches to causal inference with long-standing interrupted time series methods of analysis, its incorporation of Bayesian statistical inference, and its thorough treatment of threats to the internal, external, and construct validity of time series analyses."
--Kenneth C. Land, John Franklin Crowell Professor Emeritus of Sociology, Duke University

"Design and Analysis of Time Series Experiments gives a new generation of behavioral, biomedical, and social scientists a comprehensive resource for understanding time series issues of causality, validity, and experimental design. Drawing on time series data about social problems such as homicides, public drunkenness, speeding crackdowns, and self-injurious behaviour, the authors present complex statistical concepts and issues in an exceptionally accessible style that will surely make it a favourite, go-to book for analysts for many years to come."
--Lorraine Mazzerole, Professor, School of Social Science, University of Queensland, Australia

"Time-series designs are critically important for rigorous scientific evaluation of the whole panoply of laws, public policies, and structural interventions affecting a society's health and well-being. Earlier work on applied time series analysis by McCleary and McDowall shaped a whole generation of scientists and evaluators and led to a burst of important policy-relevant research. Their timely new text advances these methods to next level and will similarly lead to significant improvements the quality of the science."
--Alexander C. Wagenaar, Research Professor, Rollins School of Public Health, Emory University and Professor Emeritus, University of Florida College of Medicine

"The must-have book for established epidemiologists and trainees alike aiming to understand ARIMA modeling and use transfer functions to estimate the causal impact of interventions. A strong focus on study design, model diagnosis, and potential threats to validity. Highly useable and practical-a generous contribution, one you will keep near your desk and use often."
--Douglas J. Wiebe, Department of Biostatistics and Epidemiology, University of Pennsylvania

"Researchers who wade into time series analysis regard the work of McCleary and McDowall as essential reading. In both their instructional and applied work, they possess an uncanny ability to intelligibly render time series logic, estimation, and diagnosis. McCleary and McDowall have now teamed up with Bradley Bartos, and not only have they written a more comprehensive book than their prior efforts, they have done so with characteristic clarity. There is much to gain from a close reading of Design and Analysis of Time Series Experiments, by students and seasoned researchers alike."
--Robert J. Apel, School of Criminal Justice, Rutgers University

"Design and Analysis of Time Series Experiments is the long awaited follow up to McDowall and McCleary's previous works. While other resources have become available over the years, few applied statistical texts ever take time to carefully illustrate the key design principles central to the statistical method in such a clear and accessible way. Previous works were essential reading for any applied evaluation researcher. I have no doubt that Design and Analysis of Time Series Experiments will follow in this tradition."
--David K. Humphreys, Associate Professor of Evidence-Based Intervention and Policy Evaluation, University of Oxford

"There are two ways to use computers, either as an autopilot or as power steering. The autopilot is fine, if you know exactly where you are heading; but it gets you in trouble if you don't. The goal of statistical analysis is the search for patterns in data, and one of the best pattern recognition devices sits between your ears. The mathematical aspects of statistics should be in support of this endeavor, helping to tease patterns out of noisy data. This book gives concrete examples of how this is done, not by rote formulas but by conscious selection of appropriate techniques."
--Michael D. Maltz, Information and Decision Sciences, University of Illinois

"One obstacle to the use of time series experiments is that results are difficult to interpret causally. This book shows how to design a time series experiment to accurately assess the causal effects of large-scale interventions. It will do much to popularize the use of time series designs in community-wide medical trials. That can greatly expand our ability to evaluate treatment and prevention intervention programs for policy purposes, a key goal for cost-sensitive health care services."
--William R. McFarlane, M.D., Professor of Psychiatry, Tufts University School of Medicine

"Design and Analysis of Time Series Experiments is the kind of book that I not only wanted to have in graduate school but needed to have in graduate school. Fortunately for me now, and the many students who are learning time series today, they have the book from the masters."
--Alex R. Piquero, Ashbel Smith Professor of Criminology, University of Texas at Dallas

"Design and Analysis of Time Series Experiments nicely covers some important but often over looked issues, especially the identifiability of causal models. I like that it covers exploratory data analysis as a type 1 experiment. Overall, the book is very well done and will make a good senior or graduate- level course."
--Gentry White, Senior Lecturer, School of Mathematical Sciences, Queensland University of Technology

Reseña del editor

Design and Analysis of Time Series Experiments presents the elements of statistical time series analysis while also addressing recent developments in research design and causal modeling. A distinguishing feature of the book is its integration of design and analysis of time series experiments.

Drawing examples from criminology, economics, education, pharmacology, public policy, program evaluation, public health, and psychology, Design and Analysis of Time Series Experiments is addressed to researchers and graduate students in a wide range of behavioral, biomedical and social sciences. Readers learn not only how-to skills but, also the underlying rationales for the design features and the analytical methods. ARIMA algebra, Box-Jenkins-Tiao models and model-building strategies, forecasting, and Box-Tiao impact models are developed in separate chapters. The presentation of the models and model-building assumes only exposure to an introductory statistics course, with more difficult mathematical material relegated to appendices. Separate chapters cover threats to statistical conclusion validity, internal validity, construct validity, and external validity with an emphasis on how these threats arise in time series experiments. Design structures for controlling the threats are presented and illustrated through examples. The chapters on statistical conclusion validity and internal validity introduce Bayesian methods, counterfactual causality and synthetic control group designs.

Building on the earlier of the authors, Design and Analysis of Time Series Experiments includes more recent developments in modeling, and considers design issues in greater detail than any existing work. Additionally, the book appeals to those who want to conduct or interpret time series experiments, as well as to those interested in research designs for causal inference.