We are interested in applying statistical methods to the study of aggressive social interactions. In one study, we developed methods for determining from all occurrence behavioral data when pairs of mice resolve their dominant-subordinate relationship. In the image you can see when individuals exhibit dominant (red) or subordinate (blue) behavior towards each other over sessions of 20 minutes across five days. We applied Kleinberg’s burst detection algorithm to identify individual bursts of aggression and/or subordination. Within each burst we determined if animals could be described as dominant or subordinate using phi-coefficients. This allowed us to temporally plot the emergence of dominant-subordinate relationships. We also examined Markov transitions and adapted pairwise-correlation methods to assess how dominant and subordinate animals dissociate as relationships resolve in terms of their sequential patterns of social behavior.

social dynamics

In collaboration with the research group of Dr Tian Zheng, Columbia University, we have also examined the temporal sequences of aggressive interactions in groups of mice. By applying network point process models with latent ranks we are able to model features such as winner effects, bursting and pair-flips. We can best model these data using Markov-Modulated Hawkes processes and suggest that these models are suitable for analyzing social interaction event dynamics of various types.

Related Publications

Lee W, Fu J, Bouwman N, Farago P, Curley JP. 2019, Temporal microstructure of dyadic social behavior during relationship formation in mice. PLoS ONE.14(12): e0220596. PDF   Online Article

Ward OG, Wu J, Zheng T, Smith AL, Curley JP, 2022, Network Hawkes Process Models for Exploring Latent Hierarchy in Social Animal Interactions, Journal of the Royal Statistical Society: Series C. PDF   Online Article

Wu J, Ward OG, Curley JP, Zheng T., 2022, Markov-Modulated Hawkes Processes for Modeling Sporadic and Bursty Event Occurrences in Social Interactions. Annals of Applied Statistics.1903.03223. PDF   Online Article