in biweight_midvar, if more than half of the datapoint have the same value the np.median(abs(data - center) will be 0
Then it creates nan values when dividing n * num/dem because dem is 0
It should not be possible to divide by zero here, need to add an error message or divide by something else than 0