Distribution of empirical p-values of the 2019 greedy-paris clusters using three different null models: Snowballing sampling, edge shuffle, and random clusters.

<p>All figures share the same y-axis which is also on a log scale. The proportion of clusters with <i>p</i> < 0.05 in each model is 6%, 48%, and 39%, for snowballing, edge-shuffle, and random clusters respectively. Snowballing has 70% of its clusters with <i>p</i> = 1.00, whereas the other two models have zero clusters falling into this category.</p> <p>(PDF)</p>