Longitudinal Data Analysis

A Practical Guide for Researchers in Aging, Health, and Social Sciences

Jason Newsom, Richard N. Jones, Scott M. Hofer, Jason Newsom, Richard N. Jones, Scott M. Hofer

• Publisher: Routledge
• Year: 2013
• ISBN:
  9781136705465
• Edition: 1
• Topic: Psychology
• Subtopic: Research & Methodology in Psychology
• Index: Psychology

This book provides accessible treatment to state-of-the-art approaches to analyzing longitudinal studies. Comprehensive coverage of the most popular analysis tools allows readers to pick and choose the techniques that best fit their research. The analyses are illustrated with examples from major longitudinal data sets including practical information about their content and design. Illustrations from popular software packages offer tips on how to interpret the results. Each chapter features suggested readings for additional study and a list of articles that further illustrate how to implement the analysis and report the results. Syntax examples for several software packages for each of the chapter examples are provided at www.psypress.com/longitudinal-data-analysis.

Although many of the examples address health or social science questions related to aging, readers from other disciplines will find the analyses relevant to their work. In addition to demonstrating statistical analysis of longitudinal data, the book shows how to interpret and analyze the results within the context of the research design. The methods covered in this book are applicable to a range of applied problems including short- to long-term longitudinal studies using a range of sample sizes.

The book provides non-technical, practical introductions to the concepts and issues relevant to longitudinal analysis. Topics include use of publicly available data sets, weighting and adjusting for complex sampling designs with longitudinal studies, missing data and attrition, measurement issues related to longitudinal research, the use of ANOVA and regression for average change over time, mediation analysis, growth curve models, basic and advanced structural equation models, and survival analysis.

An ideal supplement for graduate level courses on data analysis and/or longitudinal modeling taught in psychology, gerontology, public health, human development, family studies, medicine, sociology, social work, and other behavioral, social, and health sciences, this multidisciplinary book will also appeal to researchers in these fields.

1

Existing Longitudinal Data Sets for the Study of Health and Social Aspects of Aging

Nathalie Huguet

Portland State University

Shayna D. Cunningham

Sociometrics Corporation

Jason T. Newsom

Portland State University

Over the past several decades, an increasing number of longitudinal data sets have become publicly available that would be of interest to researchers studying aging (Boslaugh, 2007). As an interdisciplinary field, gerontology is the study of health and social aspects of aging. Many of these studies represent the culmination of a major investment of resources and the collaboration of multiple agencies or institutions, and they represent an unprecedented opportunity for researchers. These data sets have strengths that cannot be matched with the primary collection efforts of individual researchers, such as a wider array of measures, larger or more representative samples, or data that are difficult or expensive to obtain (e.g., clinical measures or death records). There are always specific aims in mind for these large studies, but, once a data set exists, it can be fodder for the investigation of hypotheses that reach well beyond the original study aims. Recognition of the potential knowledge still to be gained from existing studies is evident at the National Institutes of Health (NIH), for example, which has issued a number of Funding Opportunity Announcements (FOAs) specifically designed to encourage secondary data analysis (e.g., National Institutes of Health [NIH], 2008a, 2008b, 2009). The availability of such rich data sources to students, junior faculty, or researchers, who might otherwise lack the time or resources needed for large-scale primary data collection efforts, provides an invaluable opportunity for career development and scientific contribution. In this chapter, we review a variety of existing longitudinal data sets that are a potential wealth of information about health and social aspects of aging. Our goal is to familiarize researchers with the key characteristics of these studies so that they may identify and obtain data from those that are most appropriate for investigating their research questions.

We begin the chapter with a discussion of where to access data from longitudinal studies on health and social aspects of aging and some important considerations when using such data for secondary analysis. This is followed by an in-depth description of several major studies including information about the purpose of the study, its design, sample size, whether mortality data are available, a general description of the measures assessed, issues related to access of the data, examples of published articles from the study, sources for obtaining the data files, and where to find documentation for the data set. At the end of the chapter, we provide a table that summarizes the key features of a more extensive list of relevant studies.

Accessing Study Data

Data are becoming more frequently available from a variety of sources, including repositories and individual researchers. Making data available has recently begun to be formalized in several policy pronouncements. The Freedom of Information Act (FOIA) mandates that investigators comply with requests for the release of their scientific data when a project has been supported by federal funds (Melichar, Evans, & Bachrach, 2002). Likewise, NIH policy stipulates, “Data should be made as widely and freely available as possible while safeguarding the privacy of participants, and protecting confidential and proprietary data” (NIH, 2003). Since 2003, investigators submitting a research application to NIH that requests $500,000 or more in direct costs in any single year must include a plan for sharing final research data, or state why data sharing is not possible (NIH, 2003).

There are several mechanisms by which investigators may make study data available. Some investigators opt to send their study data directly to requesters or post the data on their institutional or personal Web sites. In lieu of, or in addition to, this option, investigators may transfer their data to a data archive facility that will check, clean, and process the data for dissemination according to common standards and set criteria. In addition to providing a central location for researchers to access existing study data, data archive facilities often provide technical assistance for users seeking help with analyses (Melichar et al., 2002).

Examples of data archive facilities that maintain collections of longitudinal data related to health and social aspects of aging are the Inter-University Consortium for political and Social Research’s (ICPSR) National Archive of Computerized Data on Aging (NACDA) (www.icpsr.umich.edu/NACDA), the Council of European Social Science Data Archives (CESSDA) (www.cessda.org), and Sociometrics Corporation’s Data Archive of Social Research on Aging (www.socio.com).

NACDA’s mission is “to advance research on aging by helping researchers to profit from the under-exploited potential of a broad range of data sets” (the ICPSR, n.d.). Toward this end, NACDA houses the world’s largest collection of data sets from longitudinal studies on health and social aspects of aging. Each study file contains ASCII text data and documentation files and many also contain setup files and portable files for SAS, SPSS, and Stata. Data sets included in NACDA are available for free to affiliates of the ICPSR member institutions or may be purchased by non-members on a fee basis. Many (700) universities and research centers across the world are the ICPSR members. A list of the members can be found on the ICPSR Web site.

The Council of European Social Science Data Archives is a group of organizations for social science data archives across Europe. Twenty European countries are members and provide data. Access to the data varies depending on the data set. Many are downloadable but some require registration with the organization from which the data originate. The data sets are available in SAS, SPSS, Stata, and ASCII format.

Sociometrics’ Data Archive of Social Research on Aging is a smaller collection comprising only data sets from studies that are selected by a National Advisory Panel of experts in the field based on their technical quality, substantive importance to the field, policy relevance, potential for secondary analysis, and disciplinary balance. All of Sociometrics’ archived data sets come in a standardized format that includes a user’s guide, raw data file (ASCII text data), SPSS and SAS program statements and portable files, a comprehensive data dictionary, description of the frequency distribution of variables, and the original data collection instruments and codebooks. In addition, they are integrated into the Multivariate Interactive Data Analysis System (MIDAS), an online statistical analysis and search and retrieval tool that incorporates archive-level, study-level, and variable-level links to metadata documentation such as original instruments, codebooks, methodology reports, and guides. Data sets in the Data Archive of Social Research on Aging are available for free to affiliates of institutions who subscribe to Sociometrics’ Social Science Electronic Data Library (SSEDL) or may be purchased individually.

Regardless of how data are shared, investigators are required to honor the terms under which research participants gave their informed consent. As such, researchers seeking access to some longitudinal data sets related to health and social aspects of aging may be required to enter into a data-sharing agreement (also known as a licensing or data distribution agreements). Such agreements generally include criteria for data access, whether or not there are any conditions on research use, and requirements to ensure research participants’ privacy and data confidentiality. In some cases, more than one version of the same data set may be available. For example, a subset of the data set may be available for public use without restrictions, but a data-sharing agreement may be required if access to more sensitive data is desired. Investigators for whom data-sharing agreements do not provide adequate safeguards to protect human subjects may choose to make data available through a data enclave. Data enclaves or Research Data Centers (RDC) are tightly controlled, secure environments to which authorized researchers must come in order to perform analyses. Table 1.1 lists examples of such centers across the United States (U.S.) supported by the National Institute on Aging (NIA) (Population Reference Bureau [PRB], 2007). In addition to these RDCs, the U.S. Bureau of the Census in the Department of Commerce manages nine Census RDCs. Also, the National Center for Health Statistics (NCHS) and the Center Disease Control and Prevention (CDC) within the Department of Health and Human Services (DHHS), manage one NCHS RDC. Some RDCs will provide remote access to data with limitations. For example, researchers may send SAS/SPSS code via email to the NCHS RDC, and receive their output via email, but will not be able to see the individual data records.

TABLE 1.1 NIA-Sponsored Data Enclaves

Note: Not all of the data sets collected or maintained by these centers require users to conduct analyses on site.

Data Documentation Initiative

Good technical documentation, such as a “data dictionary” or “codebook,” is essential to ensure the proper use of existing data sets. The documentation accompanying a data set should include information about the following: data collection procedures, structure of the data sets, variables and values including coding and classification schemes, derived variables, weighting, other data sources, and description of the project, including the principal investigator(s), title, funding sources, specific aims, project history, geographic location of data collection, and time period covered (the ICPSR, 2005). The exact contents of the documentation will vary by study. The quality of such documentation may also vary considerably.

The Data Documentation Initiative (DDI) is an international effort to establish a standard for technical documentation describing social science data. One of DDI’s primary goals is to provide researchers with a method and language for documenting their research in a codebook form that utilizes a common structure called eXtensible Mark-up Language (XML). The DDI specification has established codebook-specific document type definitions (DTD) that encode metadata with XML “tags” similar to those found in the marked text of an HTML file. The DTD XML format has a number of qualities that enable effective, efficient, and accurate use of a data set. First, the XML documents may be directly inputed into software packages. Second, information may be displayed in any format or style the user desires and is not tied to a particular word processing or other application for display. Third, complex and precise searches may be performed on the XML documents. Last, the information can be presented as in a traditional codebook formatting (Miller & Vardigan, 2005). The flexibility that XML affords for data preparation has enormous implications for future preservation and sharing of all types of data. As more organizations and institutions provide web-based services for accessing data sets, the implementation of methods for standardized storage of data and documentation has become crucial for effective use of data.

Important Considerations Associated With the Use of Existing Data Sets

There are a number of important factors to consider when identifying data sets to use for secondary analyses. Mo...