University of Bristol

Academic Institute

Apostolos GkatzionisProfessor Kate TillingDr Kate NorthstoneDr Jon Heron

B3838

Biomedical research is often hindered by the presence of missing data. For example, missing data can occur due to study participants' unwillingness to disclose sensitive information about themselves (e.g. refusing to answer questions related to their mental health, alcohol consumption or drug use). In our research, we develop novel statistical methodologies to account for missing data, using available information on traits that affect a participant's willingness to provide full data but not otherwise affecting the outcome of an applied study. We hope to illustrate our method by using the ALSPAC dataset to estimate the true prevalence of alcoholism, depression, smoking and self-harm, as well as assessing the effects of obesity and education on these traits.

This work will showcase our new method to adjust for selection bias. If our method proves to be effective, its use will aid researchers working on applications where selection bias is suspected.

02/08/2021