The data set was downloaded using a crawling agent that obtained the complete public profile information (including personal information and the list of Facebook friends), and the Facebook wall data of a set of users in a large regional network of Facebook. The agent followed the friendship links to obtain a large connected component of the regional network. The 44% of profiles in the regional network that was not been downloaded were profiles with restrictive privacy settings or users disconnected from the giant component. Despite the high number of missing profiles, some of their data is still present in the data set. In fact, if a public profile of a user A was connected to a non-public profile B, the posts sent from B to A were still visible in A's Facebook wall. Moreover, B would appear in the friend list of A. Therefore, information exchanged on missing links from non-public profiles to public profiles is still available. We miss information related to posts (i) from public profiles (node A in our example) to non-public profiles (node B) and (ii) between non-public profiles. We discuss below how we estimate traffic related to (i). As for (ii), the amount of data collected for non-public profiles is usually lower than that of public profiles since their communication traces appear only indirectly inside the walls of other public users. For this reason, most private profiles appear as users with low Facebook usage, which we discard in our analysis. Given this, we argue that missing information about their mutual interaction is not particularly problematic for our purposes. Hence, we reasonably assume that, despite not containing all the possible communication records between users in the regional network, the data set is still a valid representation of Facebook social network for the purpose of ego network analysis.