|
Abstract:
|
Repeated interpersonal interactions over time are often treated as a multivariate point process, and network data including time of the event can be analyzed effectively using a Cox multiplicative intensity model as in Perry and Wolfe (2013). Understanding the unobserved Euclidean "social space" is another method to study network data, as in the latent space model of Hoff et al. (2002). This work introduces a new approach to use both temporal and socio-spatial components of network data, by proposing a joint model of the point process and latent positions. Estimation procedures are discussed, including maximum likelihood and Bayesian frameworks. The applicability and interpretability of the model is illustrated using the email data for county government managers from 17 counties in North Carolina.
|
