Report on methods for complex linked data

Abstract

The UK's longitudinal study resources have been largely survey-based, but there is potential for increasing the range of variables and coverage of the information through linkage and harmonisation with other datasets. Combining multiple sources in this way creates data with complex structures which require appropriate methodologies for analysis. This report describes the nature of complexities in linked datasets for analysis, and summarises the methodological requirements for:
analysis of partially overlapping repeated measures;
analysis of networks within longitudinal data;
secondary analysis of linked data;
longitudinal population size estimation.
These approaches all share the feature that they have to deal with the potential for errors in the data linkage process, particularly where automated solutions are needed to control costs. A summary of the challenges in providing a scalable linkage methodology which can deal with multiple datasets is included. Secondary analysis of data that cannot be linked without errors is a central topic area in the landscape created by longitudinal data linkage.
Key areas where methodological development seems possible and useful are in the use of structural equation models and related approaches to make the best use of the all the available data, and deployment of the entity resolution approach to data linkage to deal with conflicting information in multiple sources.

More information

Identifiers

ORCID for Li-Chun Zhang: orcid.org/0000-0002-3944-9484

Catalogue record

Export record