..just an FYI, but if using 'with print_progress = 1' as suggested by Rob you may see mentions of tablescans being used to gather the non-lead indexed column stats (when not using hash based stats gathering).
This tablescan info from print_progress is bogus and as Kevin mentions it will use the smallest index to gather those stats. This should be fixed in a future release in the not too distant future, CR is 723460.
In terms of how it determines the smallest index, before 15.7 ESD#2, it was loosely based around the length of the index key when the column existed as a minor column in multiple indices.
From ESD#2 onwards it takes look at the size of indices in terms of number of leaf pages.