A holistic statistical approach for determining the relationships between social, economic and health markers using the English Longitudinal Study...

It has long been recognized that the demographics of developed societies such as Britain are undergoing a fundamental change, with people living longer lives, and as a consequence experiencing a longer period of frailty in old age. This has immediate implications for economy and labour force dynamics, as well as the social welfare system (Tinker, 2002). Increasing efforts are being made to improve our understanding of this new phenomenon. As the population, by definition, is a mixture of very heterogeneous entities, it is vital to characterise the varying experiences of ageing in order to understand the needs of individuals and to quantify the impacts of any future policy changes.
In parallel, due to advancement in technology, there has been an explosive growth in data collection, which could be utilised to improve our understanding of the ageing process on a finer scale and to aid both individual and policy-level decision making. However, these technological advancements also bring new challenges to analyse such rich information, due to the problem known as the "curse of dimensionality". This states that any standard analysis becomes a non-trivial task in a high dimensional context (many observed variables). Additionally, even in the era of big data, it is impossible to collect all the relevant information that characterises individual variability, and variability and noise levels may even increase with increasing number of measurements.