This is essentially a re-opening of #6844 with examples:

import numpy as np
print(np.__version__)  # 1.11.1
x = np.ma.arange(24)
x.shape = (4, 3, 2)
# e.g. this works
print(np.ma.mean(x, axis=(0,1)))  # [11.0 12.0]
# but it doesn't work with a mask array set
x.mask = x.data > 12

Testing with version 1.11.1, here are the methods that raise TypeError: tuple indices must be integers, not tuple:

• np.ma.mean(x, axis=(0,1))
• np.ma.std(x, axis=(0, 1))
• np.ma.var(x, axis=(0, 1))

This raises a variation TypeError: an integer is required for the axis:

• np.ma.median(x, axis=(0, 1))

And these methods appear to work, returning a masked array result:

• np.ma.min(x, axis=(0, 1)) # [0 1]
• np.ma.max(x, axis=(0, 1)) # [12 11]
• np.ma.sum(x, axis=(0, 1)) # [42 36]

(are there other operators with a mask= argument?)

I get the same results with NumPy 1.11.1 with Python 2.7 on Windows and Linux.