Skip to main content
Figure 13 | EPJ Data Science

Figure 13

From: The shocklet transform: a decomposition method for the identification of local, mechanism-driven dynamics in sociotechnical time series

Figure 13

Social dynamics in cusp segments. Extracted cusp Extracted shock segments show diverse behavior corresponding to divergent social dynamics. We extract “important” shock segments (those that breach the top \(k=20\) ranked weighted shock indicator at least once during the decade under study) and normalize them as described in Sect. 2. We then find the densities of shock points \(t^{*}_{1}\), measured using the maxima of the within-window time series, and alternatively measured using the maxima of the (relative) shock indicator function. We calculate relative maxima of these distributions and spatially-average shock segments whose maxima were closest to these relative maxima; we display these mean shock segments along with sample shock segments that are close to these mean shock segments in norm. We introduce a classification scheme for shock dynamics: Type I (panel (A)) dynamics are those that display slow buildup and fast relaxation; Type II (panel (B)) dynamics, conversely, display fast (shock-like) buildup and slow relaxation; and Type III (panel (C)) dynamics are relatively symmetric. Overall, we find that Type III dynamics are most common (40.9%) among words that breach the top \(k=20\) ranked weighted shock indicator function, while Type II are second-most common (36.4%), followed by Type I (22.7%)

Back to article page