Multiple imputation using auxiliary imputation variables that only predict missingness can increase bias due to data missing not at random
When using MI, the naïve and commonly used strategy of including all available auxiliary variables should be avoided. We recommend including the variables most predictive of the partially observed variable as auxiliary variables, where these can be identified through consideration of the plausible casual diagrams and missingness mechanisms, as well as data exploration (noting that associations with the partially observed variable in the complete records may be distorted due to selection bias).
Authors
- Jon Heron (ORCID: https://orcid.org/0000-0001-6199-5644)
- Kate M. Tilling (ORCID: https://orcid.org/0000-0002-1010-8926)
- Elinor Curnow (ORCID: https://orcid.org/0000-0002-3109-3647)
- James R. Carpenter (ORCID: https://orcid.org/0000-0003-3890-6206)
- Rosie Cornish (ORCID: https://orcid.org/0000-0002-2874-7646)
Institutions
- University of London (GB)
- University of Bristol (GB)
- London School of Hygiene & Tropical Medicine (GB)
- University College London (GB)
- Medical Research Council (GB)
Publication Details
- Journal
- BMC Medical Research Methodology
- Published
- 2024-10-07
- DOI
- https://doi.org/10.1186/s12874-024-02353-9
- Citations
- 7
- Primary Topic
- Health, Environment, Cognitive Aging
- Type
- article
- Field-Weighted Citation Impact
- 2.20
Funders
- University of Bristol
- Medical Research Council