Predicted relative errors.

Upscaling results for the number of species of the four analysed datasets from local samples covering a fraction p* = 5% of the corresponding global dataset. For each human activity, we display the number of species (users, hashtags, words) and individuals (sent mails, posts, occurrences) at the global scale together with the average fitted RSA distribution parameters at the sampled scale and the relative percentage error (mean and standard deviation among 100 trials) between the true number of species and the one predicted by our framework. See S1 Fig in S1 Appendix for the corresponding fitting curves and predicted global RSA patterns.