📘 What This Paper Does

Introduces Dynamic Network Logistic Regression (a special case of Temporal Exponential Random Graph Models) as a framework to implement decision‑theoretic models of network dynamics in panel data. Practical heuristics for model building and assessment are provided, and the approach is illustrated with a longitudinal blog citation network sampled during the 2004 U.S. presidential election cycle.

📊 Tracking Blog Links Across the 2004 Campaign

Uses a longitudinal sample of all Democratic National Convention and Republican National Convention–designated blog citation networks collected during the 2004 campaign. The case is highlighted because it captures the institutional emergence of Internet‑based media—blogs and social networking sites—as recognized features of the American political landscape.

🔬 How the Analysis Works

Applies Dynamic Network Logistic Regression techniques to model the binary choice of whether one blog cites another over time. The analytic strategy and model assessment combine likelihood-based selection and simulation checks. The analysis tests competing mechanisms thought to shape citation choices, including:

• strategic mechanisms
• institutional mechanisms
• balance‑theoretic mechanisms
• exogenous influences such as seasonality and discrete political events

Practical guidance is offered on model specification, selection, and adequacy assessment in panel network settings.

📈 Key Results and Model Evaluation

Deviance‑based model selection criteria and simulation‑based model adequacy tests are used to compare specifications and to identify the combination of processes that best characterizes blog citation choice behavior over time. The approach demonstrates how dynamic logistic‑choice models can recover the mix of endogenous and exogenous influences on link formation in temporal network data.

💡 Why It Matters

The paper shows how a decision‑theoretic, panel‑data approach to dynamic networks can be operationalized and assessed in practice. The method is directly applicable to studying political communication in online media and other settings where actors repeatedly choose links over time.