Tests for equal dispersion (“spread”) in two or more groups of multivariate data (PERMDISP; Anderson 2005).
The procedure starts from a standard Principal Coordinates Analysis of the complete data set (all groups), using the selected distance measure. In the PCoA space, calculate the Euclidean distance zi from each point i to its group centroid. These distances are subjected to a standard one-way, univariate ANOVA. The significance (p value) is estimated by a permutation test.
PAST also includes pairwise tests across groups. These tests are based on the original PCoA of the complete data set (not PCoA on the reduced data set with only two groups).
A box plot of dispersions is provided, showing the z values for each group.
Reference
Anderson, M.J. 2005. Distance-based tests for homogeneity of multivariate dispersions. Biometrics 62:245-253.