Murdoch University Research Repository

Welcome to the Murdoch University Research Repository

The Murdoch University Research Repository is an open access digital collection of research
created by Murdoch University staff, researchers and postgraduate students.

Learn more

Snapshot or movie: How sampling methods bias dolphin social network metrics

Stanton, M.A., Mann, J., Bienenstock, E.J., Gibson, Q.A., Sargeant, B.L, Bejder, L. and Singh, L.O (2010) Snapshot or movie: How sampling methods bias dolphin social network metrics. In: International Network for Social Network Analysis, Sunbelt Conference, 29 June - 4 July, Trento, Italy.


Social network analysis is an increasingly popular technique for quantifying animal societies that provides important insight into studies of cooperation, disease transmission, and anthropogenic disturbance. However, the influence of behavioral sampling method on network metrics remains relatively unexplored. Most animal networks are constructed from independent sightings (surveys) of shared group membership. However, dyadic analyses indicate that animal focal follows (systematic sampling of a focal individual’s associations over longer periods of time) provide greater detail in terms of individual social variation. In this study we compare the size, density, and heterogeneity of 31 adult female bottlenose dolphin (Tursiops sp.) ego networks built using survey and focal sampling methods. This subset of well-studied females is part of a long-term bottlenose dolphin study in Shark Bay, Australia that has monitored >1200 individuals for the last 25 years. Additionally, our previous findings show that dolphin social network metrics vary significantly depending on the minimum observation threshold required for individual inclusion in the network. Restriction to individuals with more observations may increase accuracy and identify biologically relevant relationships, but potentially biases against less sociable individuals. We investigate the implications of observation threshold when constructing the above ego networks by creating multiple ego networks for each dolphin using an increasing number of observations randomly sampled from the dataset. Addressing the biases of sampling method and observation threshold is a critical first step in understanding the resulting limitations of inference and prediction.

Item Type: Conference Item
Murdoch Affiliation(s): Centre for Fish and Fisheries Research
Conference Website:
Item Control Page Item Control Page