Analysis of Variance of Multiple Causal Networks

Area of Science:

• Computational biology
• Network analysis
• Causal inference

Background:

• Constructing directed cyclic graphs (DCGs) is algorithmically complex and computationally intensive.
• Comparing multiple DCGs and identifying cross-graph causalities presents significant challenges.

Purpose of the Study:

• To propose a unified structural model for constructing and comparing multiple DCGs simultaneously.
• To develop a limited-information-based method for inferring network disparities and visualizing them.

Main Methods:

• A two-stage algorithm utilizing parallel computation for scalability.
• Limited-information-based network construction and disparity inference.
• Correspondence analysis for visualization and bootstrap method for statistical significance testing.

Main Results:

• The proposed method successfully unifies multiple DCGs into a single structural model.
• It enables simultaneous network construction and disparity inference with robust theoretical properties.
• Effectiveness demonstrated on both synthetic and real-world datasets.

Conclusions:

• The novel method efficiently constructs and compares multiple DCGs, addressing previous limitations.
• It offers a scalable and statistically robust approach for analyzing complex network relationships.
• The technique facilitates the identification of perturbational causalities across different networks.