mean value of arrays w/ UncertainQuantities #77

I have a 2D array and there is one standard deviation that describes this array, as calculated by np.std(arr). However, UncertainQuantity requires the uncertainty to be the same size as the array. I guess this makes sense since the uncertainty applies to every element, and some arrays may have different uncertainties with different elements. Still it would be convenient to be able to code like this:

x = np.array([1,2,3]) * pq.m
y = pq.UncertainQuantity( x, pq.m, np.std(x))
print y

That doesn't work because np.std(x) isn't the same size as x. My solution is to code this:

y = pq.UncertainQuantity( x, x.units, np.zeros_like(x)+np.std(x))

Now I perform some operations on the array y:

x = np.array([1,3,1,3,1,3]) * pq.m
y = pq.UncertainQuantity( x, x.units, np.zeros_like(x)+np.std(x))
print "\ny: ", y
print "\nsum(y): ", np.sum(y)
print "\nmean(y): ", np.mean(y)
print "\nnanmean(y): ", np.nanmean(y)
print "\nmymean(y): ", np.sum(y)/np.size(y)
#+END_SRC

y:  [ 1.  3.  1.  3.  1.  3.] m
+/-[ 1.  1.  1.  1.  1.  1.] m (1 sigma)

sum(y):  12.0 m
+/-2.44948974278 m (1 sigma)

mean(y):  2.0 m

nanmean(y):  2.0 m
+/-2.44948974278 m (1 sigma)

mymean(y):  2.0 m
+/-0.408248290464 m (1 sigma)

Issues seem to be

• np.mean drops the uncertainty (but np.nanmean does not)
• np.nanmean returns an incorrect result. The result from np.sum(y)/np.size(y) is correct.

Now I'm going to look into the code to see if I can easily fix this...