Curating Social Science Data The View from ICPSR

1
Curating Social Science DataThe View from ICPSR

• Myron Gutmann
• Digital Curation Conference
• December 12, 2007

2
To Start...

• Curation is important probably the most
  important thing we do
• Curation is the process that prepares data for
  use, for re-use, and for long-term preservation
• Im going to concentrate on preparation, and how
  it supports use, re-use or preservation
  themselves
• Standards are critical!
• What are we doing with social science data at
  ICPSR?
• At the end, Im going to talk a bit about the
  future, and what I think that means for data
  curation

3
ICPSR in a Minute

• Membership organization at University of Michigan
• Founded 1962
• 6,000 Studies 500,000 data files
• Social Science more, with increasing level of
  confidential data
• Funds from 600 members, government agencies
  sponsors, research grants
• Part of a Global network of social science data
  archives, with formal organizations in U.S.
  Europe

4
Data in Quantitative Social Science

• Since the 1940s
• Surveys Polls Large Small Private
  Public
• Administrative Microdata and Tabulated data
  (Census other official statistical agencies)
• Data Innovations from the 1960s 1970s
• Longitudinal data from surveys
• Longitudinal data from linked historical sources
• Magnetic tape storage eased sharing and
  preservation

5
Dramatic Changes Since the 1990s

• Most recently transactional data from cell
  phones, on-line games, blogs, EZ-Tag, etc
• Retrospectively Harmonized data
• IPUMS (Census data U.S. international)
• Surveys, including NHIS, CPES, Fertility...
• Expanding the science to include
• Biomedical environmental data
• Bibliography as data
• Web-based enclave data access

Onnella in PNAS May 2007
6
Data Preservation SharingSocial Science
Practices

• Longer and better sharing than in many sciences
• Heterogeneous tradition
• Strong in some disciplines, weak in others
• Function of use of large datasets in research
  graduate instruction
• Supported by NSF, NIH, NIJ Policies, Census
• DDI Metadata Standard
• Strong Curation Practices

7
Minding our Knitting (or what I tell young data
archivists)

• Curation matters, to ensure reuse and document
  provenance - Quality Control is essential
• Our scientific authority is critical, as
  established by a large number of reports that
  mention ICPSR (and other major data archives) as
  key infrastructure
• We must protect the human subjects of research
  who are the basis of what we do
• Social science and technology are constantly
  changing

8
Curation is Easy, Isnt it?

• Identification, appraisal, acquisition are the
  first part of the curation process
• Curation process first invented in the 1960s
  clearly in need of reinvention.
• Is curation manual or not? ICPSR employs a mix of
  manual automated processes, with ever more
  automation adherence to standards

9
Knowing the Business Process Means Knowing how to
Get the Work Done
10
Metadata MattersCritical Curation Needs

• Data Management Variable-level Metadata
• Validity
• Confidentiality
• Usability
• Data Discovery Authority Study-level Metadata
• Discovery Planning
• Provenance
• Versioning
• Preservation Planning Preservation Metadata

11
Validity Confidentiality

• Critical task are the data valid and respondents
  anonymous?
• Validity
• Compare producers documentation with data
• Identify questionable data
• Discuss correction with producer
• Document any changes
• Confidentiality
• Disclosure risk assessment reduction

12
Usability

• Critical task How easily can the data be used?
• Solution 1 Variable-level metadata
• Create complete variable-level metadata
  (including question text or equivalent)
• Create complete traditional codebooks on-line
  documentation
• Solution 2 On-line Delivery Analysis
• Effective on-line delivery systems since 2001
• On-line analysis systems for many users
• Quality Control with managerial sign-off

13
Discovery The Problem
14
Discovery Solutions

• Critical task How easily can the data be found?
• Process
• Study-level metadata creation
• High quality is critical
• Producer-supplied metadata is not sufficient
• Metadata must be effectively exposed-and
  therefore standards-based
• Participation in multi-institution portals,
  including DataPASS (US) and CESSDA (Europe)

15
Provenance, Version, Preservation

• Critical task Data users need to know what they
  are getting, to achieve replication, and the
  archive needs to know preservation history
  future needs
• Process must provide
• Provenance metadata (where did it come from?)
• Version metadata (when did it change?)
• Preservation metadata (when should we review and
  possibly change?) practice (another story)
• The Challenge How much detail is required?

16
Research Curationfor the Future
17
Great Science in the Future will Require
Massively Complex Data

• Many Science Domains
• Social Science, biomedical, environmental,
  geographical, psychological, engineering,
  transactional
• Examples
• National Childrens Study
• Chicago Neighborhoods
• Distributed Sources Replication
• Heterogeneous, not Homogeneous Metadata

18
Challenges to Curation, Preservation, Access in
the Future

• Massively Complex data means
• No single repository
• No single harmonization scheme
• Greater risk of disclosure confidentiality
  breach
• Data Arent All Digital or Quantitative
• Qualitative, including Video Audio
• Biological Samples
• Bibliography
• Living Studies that change frequently

19
Curation for the Future

• Technology that allows disparate data and
  researchers to talk, even when they dont know
  the same language
• Ability to build new dynamic communities around
  new groupings of data questions
• Engage technologists domain scientists to build
  new systems together

gutmann_at_umich.edu