Box 1 Line-up of steps

1.

Define the network edges by choosing a sensible distance threshold based on the research question, species sociality, and information obtained from the data (see section Identifying Associations and Network Construction for more details).

2.

Check if the interactions captured by the sample are non-random with the help of network permutations. Network metrics can be deemed suitable after assessing whether they capture non-random associations via network permutations (Step 1).

3.

Identify stable network metrics concerning the species and the available sample using sub-sampling from the observed data (Step 2).

4.

Identify the minimum sampling effort required to determine the network properties that are different from a randomly generated network by comparing the sub-sampled networks from permuted data sets with the sub-samples of the observed data (Step 2).

5.

Obtain confidence intervals around the point estimates of network metrics using the bootstrapping algorithm, which also takes into account the autocorrelated structure of telemetry relocation data. The width of confidence intervals can also be analysed for lowering sample sizes (Step 3).

6.

To assess which node-level network metric remains least affected with lowering sample sizes, obtain a correlation coefficient between the node-level metrics from the observed sample and the same nodes from the sub-sample. The local network metrics with a high correlation (>0.7) are expected to be more stable and should be chosen for further analysis as they are more likely to represent the position of individuals in the network, similar to their position in the full population (Step 4).

7.

To estimate the uncertainty in the observed values of node-level network metrics, obtain confidence intervals around the point estimates for each node (Step 5).