B4170 - Assessing the impact of missing data in auxiliary variables on multiple imputation estimates - 02/11/2022

B number: 
B4170
Principal applicant name: 
Paul Madley-Dowd | University of Bristol (United Kingdom)
Co-applicants: 
Professor Kate Tilling, Dr Jon Heron, Dr Rachael Hughes, Elinor Curnow
Title of project: 
Assessing the impact of missing data in auxiliary variables on multiple imputation estimates
Proposal summary: 

Data are frequently missing in observational cohorts which can lead to bias in estimates of effect sizes between exposure and outcome. We use a method called multiple imputation to try to correct for this. Auxiliary methods are often essential for removing bias in multiple imputation analyses, but are frequently missing themselves. We aim to apply simulations and an empirical example (using ALSPAC data) to assess how missing data in auxiliary variables can impact multiple imputation analyses.

Impact of research: 
We aim to improve current practice of the implementation of multiple imputation.
Date proposal received: 
Friday, 14 October, 2022
Date proposal approved: 
Monday, 24 October, 2022
Keywords: 
Statistics/methodology, Cognitive impairment, Statistical methods, Methods - e.g. cross cohort analysis, data mining, mendelian randomisation, etc.