Note that if the associationMatrix contains values of different effectsizes,
the sorting may be misleading. For example, a value of Cohen's d of .45 is higher
than a value of Pearson's r of .35, and so will end up higher in a 'decreasing'
sort - even though the association represented by an r of .35 is stronger than
that represented by a d of .45.
Furthermore, only asymmetrical associationMatrices can be sorted; sorting a
symmetrical association matrix would also change the order of the columns, after
all.