Relative importance of dissemination across users, dissemination across threads, and frequency in word dynamics.

Relative importance of the three factors as predictors of frequency change (), calculated using the method of Ref. [62]. Importance is based on the fraction of the variance of explained by each factor. This method conservatively estimates the relative importance of the independent variables in a multiple regression setting. The data are combined over all window pairs considered in Figure 3. To avoid artifactual correlations for small and large , the range of words is restricted in , as indicated in the caption of Figure 5.