Triadic patterns of friendships in YouTube groups

This paper looked for previously unexplored triadic network patterns in online social networks in general and of YouTube online friendship network in particular. In order to accomplish statistically sound analysis of local micro processes acting on triadic level in social networks, we applied statistical tests of triad census, by conditioning expected counts of triads on observed values of dyad census. This allowed for testing of arrangement of dyads in triads, which opened room for some novel discoveries and interpretations when it comes to network structure of online social networks. In order to allow for otherwise computationally too expensive analysis of large online social networks, our network analysis considered YouTube groups. Beside confirmation of high clustering, we found following: reciprocal links tend not to stand alone in triads, significant number of users tend to create links with many other users irrelevantly of whether these many others respond to their friendships or not, users who already have existing reciprocal links with some other users will tend to reciprocate incoming friendship links more than other users, there exist significantly many users who have lots of reciprocal links with other users thus acting as reciprocity hubs.