average of a number of arrays with numpy without considering zero values

Question

I am working on numpy and I have a number of arrays with the same size and shape like: a= [153 186 0 258] b=[156 136 156 0] c=[193 150 950 757] I want to have average of the arrays, but I want the program to ignore the zero values in the computation. So, the resulting array for this example will be: d=[167.333 157.333 553 507.5] this is the result of this computation: d=[(153+156+193)/3 (186+136+150)/3 (156+950)/2 (258+757)/2]. Is it possible to do that?

Answer 1

In Python:

>>> a = [153, 186, 0, 258]
>>> b = [156, 136, 156, 0]
>>> c = [193, 150, 950, 757]
>>> import statistics
>>> [statistics.mean([x for x in s if x]) for s in zip(*[a, b, c])]
[167.33333333333334, 157.33333333333334, 553, 507.5]

In numpy:

>>> import numpy as np
>>> A = np.vstack([a,b,c])
>>> np.average(A, axis=0, weights=A.astype(bool))
array([ 167.33333333,  157.33333333,  553.        ,  507.5       ])

If there is a possibility that all values in a column can equal zero, you may want to use masked arrays to avoid the problem that the normalization is impossible (weights can't sum to zero). Undefined slots in output will be masked.

>>> a[0] = b[0] = c[0] = 0
>>> A = np.vstack([a,b,c])
>>> np.ma.average(A, axis=0, weights=A.astype(bool))
masked_array(data=[--, 157.33333333333334, 553.0, 507.5],
             mask=[ True, False, False, False],
             fill_value=1e+20)
>>> np.ma.average(A, axis=0, weights=A.astype(bool)).tolist()
[None, 157.33333333333334, 553.0, 507.5]