Correcting for the impact of gregariousness in social network analyses

The social network approach provides a set of statistical tools to analyse associations between individuals. The 'half-weight index' (HWI), the association index most commonly used in social network analyses, does not take into account differences between the gregariousness of individuals. Thus, the HWI may not be a good measure of relationships between individuals: it could indicate strong affinities that do not exist and vice versa. Here we present a new index, the HWIG, that corrects the association index between two individuals for their respective levels of gregariousness. We compared the HWIG to the HWI by simulating populations in which individuals varied in their gregariousness and their affinities for each other. Unlike the HWIG, the estimation of associations made by the HWI was strongly influenced by the gregariousness of individuals: the HWI was systematically less strongly correlated with the true (input) affinity than the HWIG and this discrepancy increased when variation in individual gregariousness increased. We recommend using the HWIG, or similar variants of other common association indices, as unbiased measures of association between individuals.