Incentivising the uptake of reusable
metadata in the survey
production process
ESRA15
Reykjavik
July 2015
Louise Corti
Collections Development and
Producer Support
Why worry about metadata?
• No universal language used to document
questions and variables
• Too many bespoke systems and vocabularies
around
• Massive waste of human resource in the survey
data lifecycle
• Interoperability saves money
• Why don’t we all use the Data Documentation
Initiative (DDI)?
Who needs incentivising?
Show how to exploit metadata for surveys
• Challenge – to get established survey operations to recognise the benefits of reusable metadata
• Midlife study in the US (MIDUS) quite unique!
• Help funders, owners and producers ‘See the light’
• For this we need to show something very cool
• Some good experimental stuff happening
Benefits of publishing rich survey metadata
• Survey documentation systems
• Question banks
• Survey data exploration systems
• Nesstar
• SDA
• Bespoke visualisation systems
Published outputs – question bank
Published outputs – online access
The reality
• Hard to match up Question and Variable information
• Too much manual data entry involved in publishing
• Must do better
• Gain rich reusable metadata from the survey design and production process
Survey production lifecycle
• Beset with manual processes
• Legacy systems
• Reluctancy to change or adapt systems
• Hard to embrace new ways – disruptive,
expensive
Typical process – worst case scenario
• Manual questionnaire entry
(doc/excel/database)
• Export in word format
• Deliver to survey agency
• Manual transfer to IBM Data Collection
• Export SPSS and PDF/word questionnaire
Survey Metadata: Barriers & Opportunities
Workshop: 26 June 2014
Meeting outcomes
• Great turn out and knowledge exchange!
• Quick turn around of principles into a ‘campaign’
document and a published ‘Questionnaire profile’
• Some very positive responses – shared problem
• Be an advocate!
Increasing use of XML for survey design and
publishing
Such as:
• Social science data archive published survey
metadata (DDI 2.5)
• Essex panel studies - bespoke XML Questionnaire
Specification Language for survey design
• UK LifeStudy – survey design instrument – XML
Discussing DDI implementation today
• CLOSER cohorts portal using DDI 3.2 Questionnaire
Profile
• DASHISH DDI 3.2 use
• Blaise – import by Michigan Questionnaire
Documentation System (MQDS) DDI 3
• IBM Data Collection DDI experiments
Short brochure for sharable survey products
• Work closely with data owners and producers
• Existing information on data sharing complex
• What is really expected!
• Transferrable information
• Not a bible
Sticks?
• Specifying data documentation requirements in the
commissioning tender for fieldwork
• Mapping between questions and data outputs
• Improved readable questionnaire for end users
CLOSER project
• Funded variable/question discovery service
• Long-running birth cohorts & longitudinal studies
• Drivers for project
• Harmonisation (biomedical, socio-economic)
• Capacity building
• Data Linkage
• Impact
• Discovery
• Encourage use of existing data resources
• Tools for enhancing survey metadata
Incentives for CLOSER PIs?
• Large award to get prestigious cohort studies on board £££
• Reduce burden - enhancement work done centrally
• Survey data managers
happy to be part of peer group
rewarding to to go back and look at data
liked a shared controlled vocabulary
Received training
variable to questionnaire mappings useful
liked visibility of their study in the search platform
Forward looking survey design
• Think upfront about reusability of questionnaire metadata
• New studies – new opportunities
• Legacy work to get old messy survey design metadata into a new environment – may be worth investing in
• Can make harmonisation work so much easier – XML schema allow formal linkages of variables across time, equivalence, differences etc.
Data publishers
• Survey owners/producers - documentation online
• Question banks
• Journals - supporting data with sufficient metadata
• Use DDI 3.2 Questionnaire profile, not bespoke
schemas
Self-deposit expectations?
• Peer review of data by data centres for all data
published – includes quality of metadata
• Journals – no unified standard for data description
or documentation
• Start with minimal metadata expectations:
• data collection description
• provenance
• data description: file and variable names, labels,
• relationships between tables/files
Some tips on incentivising
• Speak a common language
• On DDI, don’t drown in detail; use existing profiles
• Start with the lowest common denominator. Baby steps
• Show value – shiny interfaces and examples!
• Provide supporting tools where possible e.g. metadata entry
• Integrate into everyday workflows and research tools
CONTACT
UK Data Service
University of Essex
Wivenhoe Park
Colchester
Essex CO4 3SQ
• ……………..…..………………………..
T +44 (0)1206 872145
