North American DDI (NADDI) Conference 2014

Many longitudinal studies of health and aging contain thousands of variables and pose particular challenges for researchers who wish to analyze the data themselves or replicate others’ research. MIDUS (Midlife in the United States) is a national longitudinal study of approximately 10,000 Americans designed to study aging as an integrated bio-psychosocial process. The study has a broad and unique blend of social, health, and biomarker data collected over several decades through a variety of modes. Recently the study has added new longitudinal cohorts that make data management, interpretation, and analysis even more challenging. A tool that allows researchers to easily create documented and citable data extracts that are directly related to their research questions would allow more time to be spent on public health research questions instead of data management. In late 2013, the United States National Institutes of Health funded MIDUS to create a DDI-based, harmonized data extraction system. This presentation is an update on progress towards creation of a DDI-based tool that will facilitate identification and harmonization of similar MIDUS variables, while enhancing the MIDUS online repository with a data extract function. This will accomplish something unprecedented: the ability to obtain customized cross-project downloads of harmonized MIDUS data that are DDI-compliant. Doing so will greatly enhance efficient and effective public use of the large longitudinal and multi-disciplinary datasets that comprise the MIDUS study. 

The Institution for Social and Policy Studies (ISPS) at Yale University and Innovations for Poverty Action (IPA) are partnering to develop a repository for research data from randomized controlled trials in the social sciences. The repository will be an expansion – and major upgrade – of the existing ISPS Data Archive. Together with Colectica, the partners are developing a software platform that leverages DDI Lifecycle, the standard for data documentation. The software structures the curation workflow, which also includes checking data for confidentiality and completeness, creating preservation formats, and reviewing and verifying code. The software will enable a seamless framework for collecting, processing, archiving, and publishing data. This data curation software system combines several off-the-shelf components with a new, open source, Web application that integrates the existing components to create a flexible data pipeline. The software helps automate parts of the data pipeline and unifies the workflow for staff. Default components include Fedora Commons, Colectica Repository, and Drupal, but the software is developed so each of these can be swapped for alternatives. This session will include a live demonstration of the data curation software. 

A DDI Alliance working group on qualitative data has developed a draft model for metadata related to qualitative data collections which also includes the possibility of hybrid collections having both qualitative and quantitative data. This paper describes two alternatives for representing categories, codes and memos from qualitative datasets. These metadata link to segments of qualitative objects. The current qualitative model includes “Analytic Category”, “Analytic Code” and “Memo” metadata elements designed to capture information from qualitative data analysis methods, as well as a “DataSet” element designed to contain related quantitative data, from procedures such as text mining. An alternative approach would be to represent codes, categories, and memos as variables in a data record, associated with a segment of a qualitative object, along with other related any quantitative data. Existing mechanisms in DDI could be used to describe these descriptive variables, as well as share them across studies. Advantages and disadvantages of each approach will be discussed.  

Earlier work by Wackerow and Hoyle has shown that DDI can be a useful medium for interchange of data and metadata among statistical packages. DDI 3.2 has new features which enhance this capability, such as the ability to use UserAttributePairs to represent custom attributes. The metadata from a statistical package can also be represented in DDI3.2 using several different styles – embedded in a StudyUnit, in a Resource Package, or in a set of Fragments. The DDI Documentation for a Fragment states “A Fragment is a means of transporting a maintainable or versionable object plus any associated notes and other material.” The DDI Documentation for a ResourcePackage states “The Resource Package is a specialized structure which is intended to hold reusable metadata outside of the structures of a single StudyUnit or Group” The DDI Documentation for a StudyUnit states “A primary packaging and publication module within DDI representing the purpose, background, development, data capture, and data products related to a study” This paper looks at metadata (including extended attributes) extracted from SAS datasets using a SAS EnterpriseGuide addin. The extracted DDI is represented as a set of Fragment elements in a FragmentInstance.