Cardiff University | Prifysgol Caerdydd ORCA
Online Research @ Cardiff 
WelshClear Cookie - decide language by browser settings

Evaluating different visualization designs for multivariate personal health data

Alrehiely, Majedah 2020. Evaluating different visualization designs for multivariate personal health data. PhD Thesis, Cardiff University.
Item availability restricted.

[img]
Preview
PDF - Accepted Post-Print Version
Download (13MB) | Preview
[img] PDF (Cardiff University Electronic Publication Form.) - Supplemental Material
Restricted to Repository staff only

Download (122kB)

Abstract

Over the past decade, self-tracking devices and apps have been increasingly used for personal data collection, particularly in the monitoring of health and physical activity. Data visualization is crucial to helping people understand and make sense of collected data. The literature review carried out on this topic revealed several studies which focus on data visualization, personal informatics systems and people’s self-tracking practice and their requirements. However, in terms of personal health data, there is a significant limitation in addressing the knowledge gap between these areas and the visualizations supported by self-tracking apps/dashboards, which people frequently use when viewing and exploring their health data. The literature review is followed by a user study and a visualization review, which explores the methods used to represent personal health data on popular self-tracking devices, apps and dashboards in order to understand the main limitations of various designs. The work goes on to address two main challenges: the need for combining multiple personal health variables to support a better representation of an individual’s health status, which could help in understanding the relationships between different health data; and to find a more suitable layout that is related to the periodic and temporal nature of personal health data. Therefore, this thesis focuses on the design and evaluation of visualizations to address these challenges by considering different design aspects of the visual layout, methods and visual encoding, while taking into account the multivariate nature of the data. With respect to the visual layout, the design proposes a novel Radial layout that utilizes an analogue clock metaphor to provide cognitive support by representing the data on an hourly basis through a clock-type display. The Radial representation is effective for data that naturally exhibit a periodic pattern. It also supports understanding of patterns with either a circular or seasonal behaviour. The developed visualizations’ evaluation process implements controlled lab-experiments. The evaluation follows a two-step method, starting with a preliminary study that leads to the design of the main evaluation study, which includes quantitative and qualitative measures of participants’ performance and preferences in the designs. The main contributions of the thesis are: (1) A thorough overview of the visualization methods provided by companion apps, dashboards and embedded displays of popular self-tracking devices. It also discusses their limitations and strengths, providing a taxonomy for the applied visualizations. (2) Suggestions and recommendations for addressing the challenges highlighted in personal health visualizations on apps and dashboards. (3) The design and implementation of multiple visualization alternatives to combine multivariate personal health data using various charts, methods and layouts. (4) Insights regarding personal health visualization gained from the results and in-depth analysis of data resulting from a controlled lab experiment. This experiment evaluated the proposed visualizations with respect to the participants’ performance regarding visual tasks related to real datasets. (5) Insights on people’s preferences on the visualizations and a structured qualitative analysis of their feedback, which demonstrated the effects of each implemented visual element. In addition to structuring design guidelines for this specific data type, the results of the study prove that the traditional Linear layout either outperforms or is comparable to the proposed Radial layout. The study also shows how the applied visual encoding and the visualization method influence the performance according to specific tasks and under different data densities. It provides plausible explanations for the significant differences in the observed performance and preferences patterns, which inform future visualization designs for personal health data.

Item Type: Thesis (PhD)
Date Type: Completion
Status: Unpublished
Schools: Computer Science & Informatics
Date of First Compliant Deposit: 9 November 2020
Last Modified: 10 Nov 2020 09:12
URI: http://orca.cardiff.ac.uk/id/eprint/136219

Actions (repository staff only)

Edit Item Edit Item

Downloads

Downloads per month over past year

View more statistics