The Challenge of Handling Structured Missingness in Integrated Data Sources
As data integration becomes ever more prevalent, a new research question that emerges is how to handle missing values that will inevitably arise in these large‐scale integrated databases? This missingness can be described as structured missingness (SM), encompassing scenarios involving multivariate missingness mechanisms and deterministic, nonrandom missingness. We highlight the challenges that this brings, especially in relation to imputation. We point out that the effectiveness of imputation depends on missingness structure, showing how the missingness structure interacts with the nature of the data. Lastly, we give guidance on handling SM in integrated data, especially in relation to imputation.
Authors
- James E. Jackson (ORCID: https://orcid.org/0000-0002-4506-7415)
- Eleni‐Rosalina Andrinopoulou (ORCID: https://orcid.org/0000-0002-5372-4163)
- Marius Hofert (ORCID: https://orcid.org/0000-0001-8009-4665)
- Sarah F. McGough (ORCID: https://orcid.org/0000-0003-2448-6714)
- Chris Harbron (ORCID: https://orcid.org/0000-0002-4353-5780)
- Niels Hagenbuch (ORCID: https://orcid.org/0000-0002-5473-075X)
- Robin Mitra (ORCID: https://orcid.org/0000-0001-9584-8044)
Institutions
- Roche (Switzerland) (CH)
- Erasmus MC (NL)
- Roche (United Kingdom) (GB)
- The Alan Turing Institute (GB)
- University College London (GB)
- University of Hong Kong (HK)
- Erasmus University Rotterdam (NL)
Publication Details
- Journal
- Advanced Intelligent Discovery
- Published
- 2025-10-10
- DOI
- https://doi.org/10.1002/aidi.202500089
- Primary Topic
- Privacy-Preserving Technologies in Data
- Type
- article
- Field-Weighted Citation Impact
- 0.00