Visualization of cause-effect relationships and trends in multidimensional time-series data.
Existing personal informatics (PI) applications help individuals to collect data related to their health and daily lifestyles but seldom offer means of analyzing collected data to find relationships and trends [4, 5]. Monitoring and analyzing personally relevant data can help people to better understand different aspects of their own lives and make improved choices over their wellbeing [1, 3, 4]. Characteristically, such data represents multidimensional time-series information since different independent variables (e.g. exercise, sleep) with multiple attributes (e.g. duration, quality) are collected over time. Visual representations of multidimensional time-series data can be effective for identifying trends, patterns and anomalies 2.
In this project, our focus is to compare and evaluate different visual representation techniques for finding relationships and trends in multidimensional time-series data. We are collaborating with our domain partner, Calgary Headache Assessment and Management Program (CHAMP), to help their chronic headache patients in exploring relationships and trends in their collected data. The project will help patients to identify triggers of headaches and make necessary adjustments in their daily lives to avoid recurring headaches. CHAMP will help us to recruit participants for the user studies component in this project.
Problem: CHAMP patients collect data daily about their headache symptoms and medications, but they cannot easily explore the effect of lifestyle parameters (e.g. food, sleep, exercises) and environmental factors (e.g. temperature, air pressure, noise) on triggering chronic headaches at different times. To enable such analysis, PI applications on mobile devices have been found to be effective for data collection and analysis due to their mobility and pervasiveness [6, 7]. Visualizing multiple time-series data on small screens (e.g. smartphone, tablet) is challenging due to the limited screen real estate. The problem can get worse if the visualizations are beyond basic charts (e.g. line graphs), since the chronic patients are not usually trained to comprehend complex temporal visualizations 6. Therefore, it is unknown what kinds of visualizations are most effective for chronic patients to perform analysis tasks in the health informatics context. Two analysis tasks are as follows. Task 1: Find probable cause (e.g. lack of sleep, weather) and effect (e.g. migraine) relationships as they happen across multiple time series events to generate hypotheses (e.g. lack of sleep as cause of migraine). Task 2: Explore past trends of selected causes based on a hypothesis (e.g. all headache episodes due to lack of sleep) to understand long term impact of the probable cause.
Research Questions: Our research aims to address the following research questions.
RQ1. What kinds of representations of multiple time-series events are appropriate for finding cause-effect relationships using small screens?
RQ2. What kinds of visualization techniques are most effective for representing sparse time events on small screens to depict trends of multiple variables?
Research Method:To evaluate the effectiveness of the visualization techniques we will apply a qualitative approach to compare different representation techniques for finding cause-effect relationships and trends. Our research will involve requirements gathering and developing a prototype.
· User Experience Scenario (Sept’14-Nov’14): We will adopt a scenario based evaluation approach described by Lam et al. 6 to evaluate our visualizations. We will conduct a user experience scenario to receive subjective feedback and opinions from CHAMP participants by showing them different alternative representations of visualization sketches for both tasks specified in the problem section. We will also introduce some existing PI applications for headache management to the patients. The goal is to understand to what extent these prototype visualizations or the existing applications support the intended tasks as seen from the participants’ perspective for requirements. We will record the perceived effectiveness of the prototype visualizations from the participants.
· Visual Data Analysis and Reasoning Scenario (Dec’15-July’15): A cross-platform touch optimized mobile PI application will be developed for chronic headache patients based on the requirements. We will deploy the application for CHAMP participants for a period of four months. We will monitor how long and how frequently patients use the visualizations for the specified tasks over this time period. At the end of study period, we will interview participants to find out how effective the visualizations are for finding cause-effect relationships to identify probable triggers, and exploring trends to understand the impact of different factors on headaches. We will compare aggregated feedback from interviews with logged data and subjective feedback from the user experience scenario to determine the effectiveness and applicability of the visualizations to perform the specified tasks.
We will produce a personal informatics application for headache patients to enable them to explore multidimensional time-series data to find triggers and see trends on mobile devices. Our research will help inform the design of visualizations for PI applications on mobile devices. The outcome will lead to better patient-clinician collaboration and discussion during appointments.
1. Li, I., Medynskiy, Y., Froehlich, J. and Larsen, J. 2012. Personal informatics in practice: improving quality of life through data. In Proc. CHI 2012 Extended Abstracts on Human Factors in Computing Systems, 2799–2802.
2. Zhao, J., Chevalier, F., Pietriga, E., and Balakrishnan, R. 2011. Exploratory Analysis of Time-Series with ChronoLenses. IEEE Transactions on Vsualization and Computer Graphics, December 2011, Vol. 17, No. 12, Pp. 2422-2431.
3. Carpendale, S., Tory, M., and Tang, A. 2014. A personal perspective on visualization and visual analytics. In Proc. of the 2014 companion publication on Designing interactive systems (DIS Companion ’14). ACM, New York, NY, USA, 223-225.
4. MacLeod, H., Tang, A., and Carpendale. S. 2013. Personal informatics in chronic illness management. In Proceedings of Graphics Interface 2013 (GI ’13). Canadian Information Processing Society, Toronto, Ont., Canada, Canada, 149-156.
5. Rapp, A., Cena, F. 2014. Self-monitoring and Technology: Challenges and Open Issues in Personal Informatics. In Proc. of HCI International, Crete, June 2014.
6. Lam, H., Bertini, E., Isenberg, P., Plaisant, C., and Carpendale, S. 2012. Empirical Studies in Information Visualization: Seven Scenarios. IEEE Transactions on Visualization and Computer Graphics, 18(9):1520–1536, September 2012.
- S. M. Waliur Rahman, Rahul Kamal Bhaskar, Frank Maurer, Anthony Tang: Supporting Chronic Headache Patients with Visual Analytics. In Proceedings of the Workshop on Personal Visual Analytics, DIS 2014, Vancouver, Canada, 2014.