🔎 What This Paper Addresses
Network analysis often examines whole-network formation while overlooking variation within and across networks — in particular, the distinct roles that actors may adopt over time. Cross-sectional methods exist for inferring latent roles, but approaches tailored to longitudinal networks remain scarce. This paper articulates the conceptual dynamics of temporally observed roles and introduces a new statistical tool to fill that gap.
🛠️ A New Method for Role Detection
The ego-TERGM is a novel estimator designed to uncover latent role dynamics within temporal networks. Key features include:
- Detects actor-level heterogeneity and latent roles as they evolve across observations
- Integrates ego-centered temporal dependence into a unified modeling framework
- Targets role classification within the context of a broader, changing network structure
⚙️ How It Is Estimated
The model is estimated via an Expectation–Maximization algorithm. This estimation strategy makes the ego-TERGM computationally efficient and practical for applied settings, enabling rapid inference about latent role membership across time points.
đź“Š Performance and Illustration
Evaluation indicates the ego-TERGM is both quick and accurate at classifying roles embedded in longitudinal network data. An application to the Kapferer strike network demonstrates the model's utility in an empirical setting and shows how role dynamics can be recovered from temporally ordered interactions.
đź’ˇ Why It Matters
By providing a method tailored to longitudinal data, the ego-TERGM allows researchers to move beyond static, whole-network summaries and to analyze how individual actors assume, switch, or sustain latent roles over time. This advance offers clearer insight into within-network heterogeneity and the processes that generate temporal network structure.
