Participant drop-out reduces longitudinal sample sizes, and when drop-out is related to variables of interest it can introduce substantial bias. Identifying predictors of drop-out can help alleviate these problems by identifying those who are most vulnerable to dropping out. This informs strategies to help better retain participants, to oversample groups of participants most likely to drop out, and to apply statistical corrections for drop-out. One commonly used statistical correction is data weighting, which seeks to correct results for drop-out by up-weighting participants with a high likelihood of being lost from a study, and down-weighting those with a low likelihood of being lost from a study.

This project is concerned with identifying the factors that predict drop-out in the large UK representative Understanding Society study. In this context, we will also evaluate the extent to which data weighting provides an adequate solution to drop-out and compare it to another widely used technique to correct for drop-out: multiple imputation. This will allow us to better understand who tends to drop-out of longitudinal studies and how this can best be addressed. The project is funded by an Understanding Society Survey Methods Fellowship awarded to Aja Murray.

Principal Investigator: Aja Murray, Department of Psychology

Co-Investigator: Tom Booth, Department of Psychology

• More about the Understanding Society study