Averaging multiple masked numpy arrays in python

Question

I have a problem that puzzles me a lot the last days. I want to average different masked arrays but it seems I cannot because my mapped arrays do not have the same map pattern.

For example. I have three arrays:

 [ --  --  --]
 [ 1.  2.  --]
 [ 2.  3.  --]
 [ --  --  --]

 [ --  2.  2.]
 [ --  2.  3.]
 [ --  --  --]
 [ --  --  --]

 [ 2   1.  --]
 [ 1   1.  --]
 [ --  --  --]
 [ --  --  --]

I want the result to be the average of these arrays, but the averaging of a masked element with a valid element should not take account the masked elements. That means in the (0,0) position I have two masked elements and 1 valid ( value =2 ), so the final result should calculate only the average of the valid elements.

 [ 2   1.5  2]
 [ 1   2    3]
 [ 2   --  --]
 [ --  --  --]

I have 28 arrays like them that I want to combine and so far I have to do complicated loops and actions to achieve the expected results. Is there any efficient way to do it?

All the arrays have the same size. Sorry It was in error in array one. — Numby
– Numby, Commented Nov 14, 2014 at 13:57
@Evert No it does not work, when I am trying for example: (array1+array2+array3).mean(axis=0) then I get only a mean for the elements that are valid in EVERY array. — Numby
– Numby, Commented Nov 14, 2014 at 14:03
Of course it does work; you just have to take the intermediate step, like Saullo explains below. — user707650
– user707650, Commented Nov 14, 2014 at 14:59

Saullo G. P. Castro · Accepted Answer · 2014-11-14 14:15:01Z

1

You can transform them into a 3-D masked array and then take the average along axis=0, for example:

np.ma.array((a, b, c)).mean(axis=0)

Example:

import numpy as np

a = np.ma.array([[99, 99, 99],
                 [ 1.,  2., 99],
                 [ 2.,  3., 99],
                 [99, 99, 99]],
                 mask=[[True, True, True],
                       [False, False, True],
                       [False, False, True],
                       [True, True, True]])
b = np.ma.array([[99,  2.,  2.],
                 [99,  2.,  3.],
                 [99, 99, 99],
                 [99, 99, 99]],
                 mask=[[True, False, False],
                       [True, False, False],
                       [True, True, True],
                       [True, True, True]])
c = np.ma.array([[2., 1., 99],
                 [1., 1., 99],
                 [99, 99, 99],
                 [99, 99, 99]],
                 mask=[[False, False, True],
                       [False, False, True],
                       [True, True, True],
                       [True, True, True]])

and:

np.ma.array((a, b, c)).mean(axis=0)

masked_array(data =
 [[2.0 1.5 2.0]
 [1.0 1.6666666666666667 3.0]
 [2.0 3.0 --]
 [-- -- --]],
             mask =
 [[False False False]
 [False False False]
 [False False  True]
 [ True  True  True]],
       fill_value = 1e+20)

answered Nov 14, 2014 at 14:15

Saullo G. P. Castro

59.4k28 gold badges191 silver badges244 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Numby Over a year ago

Thanks Saullo! That works exactly as I would like :)

Collectives™ on Stack Overflow

Averaging multiple masked numpy arrays in python

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related