Estimates of association and prevalence for measures such as health and related determinants are commonly derived from population-based studies. Inference validity depends on generalisability, however, this can be threatened by compromised participation levels – widely recognised as a considerable and growing problem. Attempts to address arising bias have typically been limited to weights based on socio-demographic characteristics which do not capture differential health and related behaviours within population sub-groups. This seminar presents a project which aims to explore and address non-response bias by way of record-linkage and application of novel methodology, with a focus on alcohol consumption in health surveys. Consideration will be given to how modifications could be made to tailor the approach to other studies.
Presented by:
Linsay Gray (University of Glasgow)
Date & time:
March 17, 2014 4:00 pm - March 17, 2014 5:30 pm
External seminars home