Selection bias in ageing studies: novel statistical methods for analyses of incomplete datasets

While some people maintain good health during ageing, others experience a profound decline in their cognitive and physical abilities.

Understanding how these abilities are related to other health, socio-economic, and demographic factors is critical for disease prevention and elderly welfare.

Unfortunately, the data from ageing studies are typically incomplete, for example, when participants drop out due to health issues.

In such a case, the probability of missing data is related to health and, therefore, to multiple partially observed health-related variables.Current statistical methods allow analyses of incomplete datasets assuming that the probability of missing data is independent of all partially observed variables but one.

Such an assumption is unlikely to be fulfilled in ageing studies, where missingness depends on multiple variables, and the analyses might provide biased conclusions about ageing.Therefore, this project aimsto develop and implement in software methods for analyses of incomplete datasets when the reason for the missingness is related to multiple partially observed variables;to use the developed methods in empirical studies of brain and cognitive ageing based on the existing data from the Betula ageing study.The development of new methods will broaden a set of studies where unbiased analyses are possible, and the implementation in the software will allow other researchers to benefit from the new methodology.

The empirical studies will advance current understanding of the determinants of cognitive well-being and age-related diseases, such as Alzheimer’s and dementia.

Finally, even though the project considers ageing studies, the methods developed in this project will find applications in a broad range of scientific fields, where the data may be incomplete.