An approximation method for improving dynamic network model fitting.

Area of Science:

• Network Science
• Computational Social Science
• Statistical Modeling

Background:

• Dynamic networks, which evolve over time, are of significant research interest.
• The separable temporal exponential-family random graph model (ERGM) is a promising approach for modeling dynamic networks.
• Fitting dynamic ERGMs can be computationally intensive, especially for large, sparse networks.

Purpose of the Study:

• To examine model fitting for dynamic networks using cross-sectional data and relationship durations under stationarity.
• To introduce a computationally efficient approximation for dynamic ERGM parameters in sparse networks.
• To evaluate the performance of the approximation method in scenarios where standard estimation methods may fail.

Main Methods:

• Developed a simple approximation for dynamic parameters in sparse networks with moderate to long relationship durations.
• Assumed stationarity and utilized independent measures of cross-sectional network structure and relationship duration.
• Investigated various tie formation and dissolution models (Bernoulli, independent-tie, dependent-tie).

Main Results:

• The proposed approximation method is most effective for sparse networks exhibiting minimal change over time.
• This approximation addresses cases where standard parameter estimation for dynamic ERGMs is challenging.
• The study analyzes the approximation's performance across different assumptions of tie dynamics.

Conclusions:

• The introduced approximation offers a viable and computationally feasible approach for fitting dynamic ERGMs to sparse network data.
• The method provides robust parameter estimation, particularly in scenarios characterized by network stability.
• This work contributes to more efficient analysis of evolving network structures.