Nearest-neighbour clusters as a novel technique for assessing group associations

When all the individuals in a social group can be easily identified, one of the simplest measures of social interaction that can be recorded is nearest neighbour identity. Many field studies use sequential scan samples of groups to build up association metrics using these nearest neighbour identities. Here, I describe a simple technique for identifying clusters of associated individuals within groups that uses nearest neighbour identity data. Using computer-generated datasets with known associations, I demonstrate that this clustering technique can be used to build data suitable for association metrics, and that it can generate comparable metrics to raw nearest neighbour data, but with much less initial data. This technique could therefore be of use where it is difficult to generate large datasets. Other situations where the technique would be useful are discussed.