Quantifying contributions of social situations to population structure
To quantify the relative contribution of each social situation to the position of individuals in the population’s social structure (the aggregate network), we calculated the Spearman’s correlation (ρ ) for each centrality measure between each social situation and the aggregate network. For example, we correlated the degree of individuals in the co-flight network with the degree of individuals in the aggregate network, the strength in the co-flight network with strength in the aggregate etc. These comparisons resulted in 9 correlations (3 indices x 3 social situations). To further determine the contribution of each social situation to the population’s social structure we asked whether the observed correlation coefficients differed from those expected by chance by comparing observed ρ values to those extracted from reference models. We created 10,000 reference (randomized) networks using node permutations (Hobson et al. 2021). In each iteration, the node IDs within each of the three social situations were permuted without replacement and the three centrality measures were calculated for each situation. By permuting only node IDs we maintained the observed network structure while breaking the relationship between social positions within and across situations. For each iteration the reference aggregate network was created by combining the permuted networks of the three social situations. The three centrality measures were then calculated for the aggregate network. We then computed the Spearman’s correlation for each centrality measure between the social situations and the aggregate network for each of the permutation iterations. We determined statistical significance by computing a p-value as the proportion of iterations in which the observed Spearman’s correlation coefficient (ρ) was larger or smaller than 95% of the ρ coefficients in the permutated data.
Analysis was conducted in R version 3.4 (R Core Team 2013). Network analysis was conducted using the ‘igraph’ R package (Csardi and Nepusz 2006) and Muxviz (De Domenico et al. 2015). Data is provided as part of the supplementary material and the analysis code is available on GitHub (https://github.com/NitikaIISc/VulturesMovementAnalysis_manuscript1).