Testing Influence of Network Structure on Team Performance Using STERGM-Based Controls

摘要

We demonstrate an approach to perform significance testing on the association between two different network-level properties, based on the observation of multiple networks over time. This approach may be applied, for instance, to evaluate how patterns of social relationships within teams are associated with team performance on different tasks. We apply this approach to understand the team processes of crews in long-duration space exploration analogs. Using data collected from crews in NASA analogs, we identify how interpersonal network patterns among crew members relate to performance on various tasks. In our significance testing, we control for complex interdependencies between network ties: structural patterns, such as reciprocity, and temporal patterns in how ties tend to form or dissolve over time. To accomplish this, Separable Temporal Exponential Random Graph Models (STERGMs) are used as a parametric approach for sampling from the null distribution, in order to calculate p-values.