When printing the median values of the input arrays and the output variance arrays, we get this for the Gaussian tests:

This is clearly wrong. 19 is way too small and 26855 is way too big.
Upon further inspection, the equation we are using was wrong (first is incorrect and current, second is correct):
