A0937
Title: MDS-based depth for mixed-type data applied to the assessment of biological age
Authors: Ignacio Cascos - Universidad Carlos III de Madrid (Spain) [presenting]
Aurea Grane Chavez - Universidad Carlos III de Madrid (Spain)
Jingye Qian - Universidad Carlos III de Madrid (Spain)
Abstract: In a mixed-type dataset, a new procedure to evaluate the centrality of an observation is introduced. This method is then used to assess the biological age of an individual, which is derived from biomarkers, medical conditions, life habits, and sociodemographic variables. These records are mixed-type, encompassing both numerical and categorical variables some of which are nonbinary. To measure the centrality of an observation within such a dataset, Gower's distance is employed between each pair of objects to build an Euclidean representation of the dataset through multidimensional scaling. Finally, classical multivariate data depth notions in such space are used. Ultimately, an individual's biological age is evaluated by finding the age that positions its records as centrally as possible among a sample of similar-aged individuals, keeping all other features constant.
