Community structure in animal social networks

Abstract

In this thesis the ideas of network analysis are applied to systems of group living animals. A method of constructing a network of associations by combining group memberships is presented. Methods of filtering the network according to association strength are discussed.

The detection and understanding of community structure within animal social networks forms an important part of this thesis. By allowing the researcher to study (and verify the statistical significance of) intermediate scale structure in the network an insight into the biological processes which may motivate the structure can be obtained.

The various methods which have been proposed to detect community structure in networks are reviewed. The use of simulated annealing to detect the structure is discussed. This technique offers the greatest sensitivity in detecting communities, making it very suitable for the detection of the subtle structures that may exist in the constructed network.

Two case studies of group living animals are considered: a population of wild guppies and a population of Galapagos Sea lions. In both systems statistically significant community structure is found. The biological processes underlying the observed structure are discussed.

In the latter part of this thesis some methods of constructing model networks with realistic community structure are discussed. Inspired by the biological aspects of the earlier part of the thesis; these offer methods of building networks in which the size, strength, and number of communities can be controlled by the researcher.