Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Visual Analysis of Multi-outcome Causal Graphs

Mengjie Fan, Jinlu Yu, Daniel Weiskopf, Nan Cao, Huai-Yu Wang, Liang Zhou

TL;DR

The paper addresses visual analysis for multi-outcome causal graphs in health research, enabling comparative analysis across diseases. It proposes a two-stage workflow: progressive visualization for single-outcome causal discovery using PC, DAG-GNN, and HCM, followed by a comparative layout and visual encodings to compare multiple outcomes. The approach is validated with benchmark datasets, case studies on UKB and CLHLS, and expert-user evaluations, demonstrating improved interpretability and cross-outcome insights. The work provides publicly available code and lays groundwork for extending causal analysis in healthcare and other domains.

Abstract

We introduce a visual analysis method for multiple causal graphs with different outcome variables, namely, multi-outcome causal graphs. Multi-outcome causal graphs are important in healthcare for understanding multimorbidity and comorbidity. To support the visual analysis, we collaborated with medical experts to devise two comparative visualization techniques at different stages of the analysis process. First, a progressive visualization method is proposed for comparing multiple state-of-the-art causal discovery algorithms. The method can handle mixed-type datasets comprising both continuous and categorical variables and assist in the creation of a fine-tuned causal graph of a single outcome. Second, a comparative graph layout technique and specialized visual encodings are devised for the quick comparison of multiple causal graphs. In our visual analysis approach, analysts start by building individual causal graphs for each outcome variable, and then, multi-outcome causal graphs are generated and visualized with our comparative technique for analyzing differences and commonalities of these causal graphs. Evaluation includes quantitative measurements on benchmark datasets, a case study with a medical expert, and expert user studies with real-world health research data.

Visual Analysis of Multi-outcome Causal Graphs

TL;DR

The paper addresses visual analysis for multi-outcome causal graphs in health research, enabling comparative analysis across diseases. It proposes a two-stage workflow: progressive visualization for single-outcome causal discovery using PC, DAG-GNN, and HCM, followed by a comparative layout and visual encodings to compare multiple outcomes. The approach is validated with benchmark datasets, case studies on UKB and CLHLS, and expert-user evaluations, demonstrating improved interpretability and cross-outcome insights. The work provides publicly available code and lays groundwork for extending causal analysis in healthcare and other domains.

Abstract

We introduce a visual analysis method for multiple causal graphs with different outcome variables, namely, multi-outcome causal graphs. Multi-outcome causal graphs are important in healthcare for understanding multimorbidity and comorbidity. To support the visual analysis, we collaborated with medical experts to devise two comparative visualization techniques at different stages of the analysis process. First, a progressive visualization method is proposed for comparing multiple state-of-the-art causal discovery algorithms. The method can handle mixed-type datasets comprising both continuous and categorical variables and assist in the creation of a fine-tuned causal graph of a single outcome. Second, a comparative graph layout technique and specialized visual encodings are devised for the quick comparison of multiple causal graphs. In our visual analysis approach, analysts start by building individual causal graphs for each outcome variable, and then, multi-outcome causal graphs are generated and visualized with our comparative technique for analyzing differences and commonalities of these causal graphs. Evaluation includes quantitative measurements on benchmark datasets, a case study with a medical expert, and expert user studies with real-world health research data.
Paper Structure (41 sections, 5 equations, 10 figures, 2 tables, 1 algorithm)

This paper contains 41 sections, 5 equations, 10 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Causal Discovery
  4. Visualization of Causal Graphs and Visual Analysis of Causal Reasoning
  5. Comparative Visualization
  6. Requirement Analysis and Method Overview
  7. Visualization of Single Causal Graph with Various Causal Discovery Methods
  8. Causal Discovery
  9. Evaluation with Benchmark Datasets
  10. Comparative Progressive Visualization of Causal Discovery Algorithms
  11. Visual Encodings of Causal Graphs
  12. Comparative Visualization of Causal Discovery Results
  13. Progressive Visualization
  14. Multidimensional Visualization of Mixed-Typed Variables
  15. Causal Effect Computation
  16. ...and 26 more sections

Figures (10)

  • Figure 1: The workflow of our visual analysis method.
  • Figure 2: Visualization of the true causal graphs (left column) and results of the three causal discovery algorithms (right column).
  • Figure 3: The user interface of our visual analysis system for single outcome graphs consists of three views: (1) Dataset and Variables Selection View allowing for dataset (a) and variables selection (b, c) and correlation analysis (d, e), (2) Single Directed Graph View allowing for graph editing (f, g), and (3) Variables Matrix View providing insight for graph editing.
  • Figure 4: Visualization of multi-outcome causal graphs with our new layout and visual mappings. Numbers on top of each subgraph are the horizontal stress of direct extraction (left of the arrow) and our method (right of the arrow), respectively.
  • Figure 5: Visual mappings for comparison of multi-outcome causal graphs.
  • ...and 5 more figures