Harmonic mean with Python Harmonic mean (using SciPy)

Harmonic mean in Python

Use Python's SciPy to find the harmonic mean. Install SciPy first.

If all the calculation targets are not values greater than 0 (positive real numbers), an exception will be thrown. Negative numbers reduce the denominator and result in a large harmonic mean. For example, if the harmonic mean of 3, -3,4 is calculated according to the formula, it will be 12. Such values cannot be used.

harmonic_mean1.py

from scipy import stats
# 1, 2,Harmonic mean of 4 ⇒ 1.71428571429
print stats.hmean([1, 2, 4])
# 1, 0.5, 2.0,Harmonic mean of 10 ⇒ 1.11111111111
print stats.hmean([1, 0.5, 2.0, 10])


#After that, an example of trying a negative value

# 3,-3,Harmonic mean of 4 ⇒ Value Error: Harmonic mean only defined if all elements greater than zero
print stats.hmean([3, -3, 4])

#When calculated according to the formula of harmonic mean ⇒ 12 values increased!?No, it is an invalid value.
print(3 / ((1 / 3.0) + (1 / -3.0) + (1 / 4.0)))

If you want to get a value like the harmonic mean even if 0 or a negative value appears

• Since it is not an accurate harmonic mean, please refer to it by judging whether this method can be applied. I made a wrapper for stats.hmean () with exception measures. Values less than or equal to 0 are calculated by replacing them with very small positive real numbers. In the example below, I replaced it with 0.0001. If all the values are 0 or less, the average is set to 0. It is unnatural to return HM_OFFSET. Since None can be handled, if the numerical value to be calculated cannot be obtained, it is set to None. This prevents the unobtainable numbers from affecting the harmonic mean value.

harmonic_mean2.py

from scipy import stats
HM_OFFSET = 0.0001
def custom_hmean(values_list):
    u"""Wrapper with exceptions to the harmonic mean function."""
    #If all the values to be calculated are None, the harmonic mean is set to None.
    if values_list[0] is None and\
       values_list == [values_list[0]] * len(values_list):
        return None

    #If all the calculated values are 0 or less, the harmonic mean is set to 0.
    if max(values_list) <= 0:
        return 0

    #If the value to be calculated is None, it is considered that there is no data, and if it is 0 or less, HM_Set to the value of OFFSET
    return stats.hmean(
        [v if v > 0 else HM_OFFSET for v in values_list if v is not None])


print custom_hmean([3, 0, 4])                   # 0.000299982501021
print custom_hmean([3, None, 4])                # 3.42857142857
print custom_hmean([0, -1, -2, -3])             # 0
print custom_hmean([0, -1, -2, 1])              # 0.000133328889037
print custom_hmean([0, 0, 1, 0])                # 0.000133328889037
print custom_hmean([None, None, 1, None])       # 1.0
print custom_hmean([None, None, None, None])    # None
print custom_hmean([0, 0, 0, 0])                # 0

Details of scipy.stats.hmean English Refer to Wikipedia for the formula of the harmonic mean ⇒ Harmonic mean Is the reciprocal of the arithmetic mean of the reciprocal