Figure 2
From: Bivalve network reveals latitudinal selectivity gradient at the end-Cretaceous mass extinction
Observed survivorship minus expected survivorship, under a model that predicts survivorship solely based on geographic range, correlates with latitude (Regression with unequal variances, P = 0.02, horizontal and vertical standard error shown).
These points are not independent because BUs contain some of the same genera, but a Mantel test between the Jaccard distance matrix and the extinction differences between those BUs indicates no correlation between the BU assemblages and their adjusted extinction percentages (P > 0.05). Above zero indicates survivorship greater than expected given geographic range. BU numbers are shown and rudist BUs are included. Average latitude of each BU was determined from the absolute latitudes of the localities in each BU. This result is robust to changes of the model parameter, global extinction risk, use of median instead of average latitude and choice of regression analysis (equal or unequal variance). Inoceramids are excluded here, but the result is not sensitive to their exclusion. If rudists are excluded the P-value increases slightly (P = 0.03).
