WinsorizedVariance¶

Variance after winsorizing extreme values — the lowest and highest Floor[f n] values are replaced by the nearest retained value, so the count is unchanged.

$ wo 'WinsorizedVariance[{1, 2, 3, 4, 5, 6, 7, 8, 9, 1000}, {0.1, 0.3}]'
40/9

Because nothing is dropped, winsorizing always leaves enough data:

$ wo 'WinsorizedVariance[{1, 2, 3}, 1/3]'
0

A matrix is reduced column by column:

$ wo 'WinsorizedVariance[{{1, 2}, {3, 4}, {5, 100}}, 1/3]'
{0, 0}