{"id":1524,"date":"2026-06-30T22:18:23","date_gmt":"2026-06-30T21:18:23","guid":{"rendered":"https:\/\/blogs.ncl.ac.uk\/nova\/?page_id=1524"},"modified":"2026-07-01T10:18:54","modified_gmt":"2026-07-01T09:18:54","slug":"visualisation-for-investigation-of-structural-missingness-application-to-the-icicle-mobility-dataset","status":"publish","type":"page","link":"https:\/\/blogs.ncl.ac.uk\/nova\/visualisation-for-investigation-of-structural-missingness-application-to-the-icicle-mobility-dataset\/","title":{"rendered":"Visualisation for Investigation of Structural Missingness: Application to the ICICLE Mobility Dataset"},"content":{"rendered":"\n

Missing data is often treated as a problem to be minimised or ignored. This PhD research argues instead that, particularly in mobility monitoring, the structure of missingness is informative and should be visualised, explored, and interpreted. The integrity and quality of datasets are essential for robust analysis and reliable decision-making. Visualisation can play a crucial role in this context, offering intuitive ways to detect, understand, and address missingness patterns, enabling researchers to better assess dataset quality.<\/p>\n\n\n\n

This project develops and evaluates exploratory visualisation approaches to support the analysis of structural missingness within longitudinal mobility monitoring health datasets. Focusing on data from patients with Parkinson\u2019s disease in the ICICLE dataset, this thesis aims to deepen understanding of missingness patterns, enabling fully informed, data-driven analyses and supporting consideration of how to deal with missing data. The design was developed with domain experts through iterative cycles of requirements gathering, prototyping, and evaluation to ensure close alignment with analytic needs.<\/p>\n\n\n\n

\"Visualisation<\/figure>\n\n\n\n

The first contribution is a state-of-the-art report (STAR) on missing data visualisation that surveys relevant literature and introduces a novel classification. The second is a user study comparing three approaches for representing two related quality metrics, such as joint and conditional missingness, evaluating how accurately and quickly users can detect and compare patterns in these quality metrics. The third contribution is a visualisation prototype designed to facilitate exploration of missingness structures and their associated quality metrics. This prototype incorporates several visualisations, including MissVisG and the MissVis plot, applied to the ICICLE dataset. The MissVis plot combines many variables into one view to compare missingness patterns, distributions, relationships, and outliers, while MissVisG provides per-variable summaries to explore missingness. Additionally, the prototype includes the Missingness Patterns plot that visualises quality metrics for missingness patterns. Finally, a multi-step evaluation was conducted, involving collaboration with domain experts, heuristic evaluations, and in-depth interviews. These demonstrated usability and analytic value, while also guiding iterative refinement. The approaches presented in this thesis not only contribute to the visualisation of missing data but also provide a foundation for further research on handling incomplete datasets.<\/p>\n\n\n\n

Team<\/h2>\n\n\n\n