Network Granger Causality with Inherent Grouping Structure.

Area of Science:

• Network analysis
• Statistical modeling
• Computational biology
• Econometrics

Background:

• Estimating high-dimensional network models is crucial for analyzing complex biological and socio-economic systems.
• Existing methods may struggle with the sparsity and group structure inherent in these networks.

Purpose of the Study:

• To develop a novel framework for learning network structures from temporal panel data.
• To address challenges in network estimation, including sparsity and node grouping.

Main Methods:

• Introduction of a group lasso regression regularization framework for network estimation.
• Examination of a thresholded variant to handle potential group misspecification.
• Establishment of norm consistency and variable selection consistency, including direction consistency.

Main Results:

• The proposed group lasso method effectively learns network structures from temporal panel data.
• Demonstrated norm and variable selection consistency of the estimation techniques.
• The thresholded variant addresses group misspecification effectively.

Conclusions:

• The developed group lasso framework provides a robust method for high-dimensional network model estimation.
• The methodology shows strong performance in simulations and practical applications in genomics and econometrics.
• This work advances the analysis of complex systems by improving network structure learning.