Meta Data Standards for Managing and Archiving Longitudinal Data: Achieving Best Practice Melanie Spallek*, Michele Haynes* & Mark Western*presented by Steven McEachern *The Institute for Social Science Research (ISSR)
ASSDA – Queensland node Brisbane Institute of Social Science Research at the University of Queensland
WHY • Cross-sectional and longitudinal data structure is different • Current meta data standards not sufficient • Great need for international standard in best practice for archiving longitudinal data
Overview • Cross-sectional studies versus longitudinal studies >different types of longitudinal studies • Major longitudinal studies archived with ASSDA • Challenges with documenting longitudinal studies • Compare meta data standards internationally • Future plans at ASSDA
Cross- sectional Longitudinal Repeated observations over time Two or more dimensional Change over time, cause-effect, shifting attitudes • Multiple variables observed at a single point in time • One- dimensional
Different types of longitudinal studies • Repeated cross-sectional studies > new sample at different points in time > represents snapshot of population at each time point > aspect of individual’s change not available • Cohort studies > group of individuals at a similar state in the life course, studied over time > problems with drop-outs • Household panels >Household as a study unit > Number of individuals can vary (move in, move out)
Major longitudinal studies archived with ASSDA • Negotiating the Life Course (NLC) > 1500 participants at wave 1 in 1996 > five waves archived so far • Australian Longitudinal Study on Women's Health (ALSWH) > three cohorts (younger, mid-aged, older) > 40,000 participants at wave 1 in 1996 > four waves archived for the younger and older cohorts and five for the mid-aged cohort
Australian Longitudinal Survey of Ageing (ALSA)> 2,087 participants at wave 1 in 1992 > seven waves archived so far • Longitudinal Surveys of Australian Youth (LSAY) > 13,613 participants at wave one in 1995 > all four waves have been archived • Longitudinal Survey of Immigrants to Australia (LSIA) >Phase 1 (three waves) and Phase 2 (two waves) have been archived Professor Mary Luszcz with the oldest ALSA participant who is 108 years old.
Meta data standards used at ASSDA • DDI2 is used for describing cross-sec and longitudinal data • coverage of DDI2 is focused on single studies, single data files, simple surveys and aggregated data files • metadata requirements for longitudinal studies differ from that of cross-sectional studies and also across types of longitudinal studies • DDI3.1 supports the description of longitudinal data, but few archives have facilitated DDI3.1 yet
Challenges • Combining Data on Same Individuals from Repeated Surveys • How do longitudinal studies name comparable variables at different surveys? • What tools are in place to easily identify variables and their comparability? • What makes a variable incomparable?
Incomparability Survey 1: Marriage improves your health Agree Disagree Survey 2: Marriage improves your health Strongly Agree StronglyDisagree
Challenges • Combining data on same individuals from repeated surveys • How do longitudinal studies name comparable variables at different surveys? • What tools are in place to easily identify variables and their comparability? • What makes a variable incomparable? • Updating longitudinal surveys
Updating Longitudinal Surveys • Additional logic check within a study participant between surveys across time • S1 S2 S3 • S1 Osteoporosis S2 Osteoporosis S3 Osteoporosis
Comparisons among International Archives • UK Data Archive’s Survey Question Bank http://surveynet.ac.uk/sqb/introduction.asp • CentERdata uses some DDI3.1 http://www.lissdata.nl/dataarchive/concepts • Other archives have not been found to address issues relating meta data for longitudinal data archiving
Future Plans at ASSDA • Website for longitudinal data archiving • Provide guidelines for data dictionary and • variable map development • Require data dictionary and variable map • with deposit of longitudinal data
Website/ Contact Australian Social Science Data Archive18 Balmain CrescentThe Australian National UniversityACTON ACT 0200 Email: email@example.com, firstname.lastname@example.org Website: www.assda.edu.auPhone: +61 2 6125 4400 Fax: +61 2 6125 0627