High quality evidence is essential to inform policy and practice. Much of this evidence comes from research studies, including longitudinal observational studies like Understanding Society. A key strength of longitudinal studies is their ability to investigate causality by making a temporal distinction between exposures and outcomes. However, they typically face problems of missing data arising through attrition and non-response. This results in a loss of power and may introduce bias. Choosing an optimal analysis strategy to minimise bias due to missing data requires an understanding of the factors associated with non-response. This also ensures that results can be interpreted with appropriate cautions.
In this project I propose carrying out an investigation into the factors associated with non-response among young people in Understanding Society using data collected on the young people themselves as well as others in their household (or former household). I will examine whether these factors differ in different population subgroups. I will then illustrate methods used to take account of non-response in the context of a specific research question – an association identified as likely to be biased due to non-response. Future research using Understanding Society data will be able to use the findings from this work to develop an analysis strategy that will minimise bias in analyses using data from the young people. This methodological project will contribute to ensuring evidence generated from the study is robust; as such, my research will have long term positive impacts on policy decisions.
