Statistical Analysis of Animal Social Networks

Social networks are pivotal in analyzing social interactions and relationships, offering insights into the structure and dynamics of both human and animal societies. Social Network Analysis (SNA) methods facilitate understanding of interactions, knowledge flow, and relational dynamics among individuals or groups. While extensively applied in human social studies, SNA s becoming crucial for studying animal societies, with significant implications for management and conservation in changing environments. This thesis addresses the methodological gaps in animal social network analysis by introducing innovative statistical techniques tailored to the challenges posed by incomplete and autocorrelated data. Traditional SNA methods often fall short in capturing the complexities of animal interactions, necessitating the development of new approaches. The goal of this thesis is to empower ecologists with limited resources to collect abundant data to derive reliable insights from the analysis of animal social networks. Chapter 1 introduces this research within the broader context of the field by providing a comprehensive overview of the methodologies employed in animal social network studies, identifying existing gaps in the literature, and discussing the implications of these gaps for current and future research. This chapter sets the stage for the subsequent chapters, establishing the significance and context of this thesis. In Chapter 2, a comprehensive five-step protocol is introduced, integrating traditional SNA methods with novel statistical techniques to quantify bias and uncertainty in network metrics. Using GPS telemetry data from five ungulate species and a validation dataset from a near-census of a sixth species, this protocol evaluates the reliability of network metrics and guides methodological choices in animal social network research.

Chapter 3 presents aniSNA, an R package designed for ecologists conducting social network analysis with their animal observation data. Built on the workflow outlined in Chapter 2, aniSNA provides user-friendly functions to generate dependable inferences and is available for download from CRAN. The chapter includes a review of recent R packages for animal SNA and a comprehensive demonstration of aniSNA’s capabilities.

In Chapter 4, an agent-based model (ABM) is developed to simulate animal observation data, highlighting the impact of sampling decisions on the accuracy and precision of network metrics. The model provides practical guidance for designing GPS-based sampling strategies, emphasizing optimal configurations for reliable inferences and suggesting effective deployment and analytical approaches.

Chapter 5 concludes the thesis by summarizing the key findings and situating them within the broader context of animal social network literature. The ecological implications of the research are discussed, along with its contributions to advancing methodologies in animal SNA. The chapter also identifies future research directions, emphasizing the need for continued exploration of animal social behavior and network dynamics.

By developing and implementing these innovative methods, this thesis aims to enhance the robustness of animal social network analysis, offering valuable tools and insights for ecologists and contributing significantly to the field of animal ecology and conservation.