Diagnostics and Visualization of Point Process Models for Event Times on a Social Network

TL;DR

This paper develops diagnostic tools and visualizations for point process models analyzing event times on social networks, enabling validation of model fit to temporal and network structures.

Contribution

It introduces systematic diagnostics and visualization methods specifically for point process models in network settings, which were previously limited in literature.

Findings

Residual and Pearson residual analysis reveal model adequacy.

Tools effectively validate models against simulated and real social interaction data.

Visualization uncovers clustering and structural patterns in event data.

Abstract

Point process models have been used to analyze interaction event times on a social network, in the hope to provides valuable insights for social science research. However, the diagnostics and visualization of the modeling results from such an analysis have received limited discussion in the literature. In this paper, we develop a systematic set of diagnostic tools and visualizations for point process models fitted to data from a network setting. We analyze the residual process and Pearson residual on the network by inspecting their structure and clustering structure. Equipped with these tools, we can validate whether a model adequately captures the temporal and/or network structures in the observed data. The utility of our approach is demonstrated using simulation studies and point process models applied to a study of animal social interactions.