Uncertainty Visualization: Concepts, Methods, and Applications in Biological Data Visualization

Daniel Weiskopf

doi:10.3389/fbinf.2022.793819

Abstract

This paper provides an overview of uncertainty visualization in general, along with specific examples of applications in bioinformatics. Starting from a processing and interaction pipeline of visualization, components are discussed that are relevant for handling and visualizing uncertainty introduced with the original data and at later stages in the pipeline, which shows the importance of making the stages of the pipeline aware of uncertainty and allowing them to propagate uncertainty. We detail concepts and methods for visual mappings of uncertainty, distinguishing between explicit and implict representations of distributions, different ways to show summary statistics, and combined or hybrid visualizations. The basic concepts are illustrated for several examples of graph visualization under uncertainty. Finally, this review paper discusses implications for the visualization of biological data and future research directions.

Highlights

IntroductionUncertainty should be considered in the context of visual data analysis and communication
Data uncertainty can seriously affect its analysis and subsequent decision-making
We address the problem of uncertainty visualization from a broader perspective, going beyond traditional statistical graphics and supporting more complex data than individual univariate distributions of data values, and linking to advanced visualization techniques

Summary

Introduction

Uncertainty should be considered in the context of visual data analysis and communication. This is well understood in many disciplines that deal with measured data. Uncertainty is not restricted to measurements but can originate from numerical error in simulations, uncertainty in devising models, or many other sources. We discuss approaches to uncertainty visualization that do not restrict themselves to error bars. We address the problem of uncertainty visualization from a broader perspective, going beyond traditional statistical graphics and supporting more complex data than individual univariate distributions of data values, and linking to advanced visualization techniques. Uncertainty visualization is difficult and considered one of the top research problems in visualization (Johnson, 2004).

Objectives

Discussion

Conclusion