B4192 - Assessing the robustness of multiple imputation strategies in practice - 14/11/2022

B number: 
B4192
Principal applicant name: 
Elinor Curnow | University of Bristol (United Kingdom)
Co-applicants: 
Professor Kate Tilling, Dr Jon Heron, Dr Rosie Cornish
Title of project: 
Assessing the robustness of multiple imputation strategies in practice
Proposal summary: 

Around 14,000 children initially participated in ALSPAC. Over time, many children have been lost to follow-up (“drop-out”) or have participated at some time-points and not at others (“sporadic non-response”). Missing data due to non-response can lead to bias when estimating associations. Multiple imputation (MI) is an analysis strategy that can correct this bias, by recovering information about the missing values using observed variables (“predictors”). However, there are methodological challenges with the correct application of MI. The aim of this project is to assess their impact in practice, and develop an automated expert system to guide researchers through the multiple decisions required when using MI.

Impact of research: 
The likely output of this research will be at least one publication in a peer-reviewed epidemiology journal, the impact of which may be improved practice by researchers when performing analyses with missing data.
Date proposal received: 
Friday, 4 November, 2022
Date proposal approved: 
Monday, 14 November, 2022
Keywords: 
Statistics/methodology, Cognitive impairment, Obesity, Statistical methods, Methods - e.g. cross cohort analysis, data mining, mendelian randomisation, etc., Statistical methods