Figure 1

Illustration of covariates related to past sentiment expression behavior of users and social contacts as well as the structure of their social network neighborhood. Nodes represent users in the social network (the gray node represents the focal user), arrows represent the follower relationships, and numbers inside of nodes represent sentiment expression history (numbers of positive and negative sentiments expressed at given time; neutral sentiments are also counted, though we do not depict them here). The direction of the arrows represents the direction of information flow. Covariates $f_1$, $f_2$, and $f_5$ are explained in the article; the remaining covariates are explained in Additional file 1. For instance, the figure indicates that $f_1$ relates to the number of followees of the focal user and $f_2$ relates to the number of tweets these followees make, whereas $f_5$ counts reciprocated follower-followee relationships. Other covariates include information about the followers (as measured by $u_2$), the number of tweets made by the user ($u_1$), triangle-based covariates that measure certain types of clustering ($f_6$ and $f_7$), and numbers of follower and followees of the user’s followees ($f_3$ and $f_4$). The figure illustrates that the values of these covariates may change with the advance of time (e.g. new tweets, new follower relationships, etc.).