weights argument seems to be ignored in np.histogram

>>> import numpy as np
>>> a=np.array([1,2,3,4,5,6,7,8,9,10])
>>> weights=np.array([0.8,0.1,0.1,0.1,0.1,0.1,0.1,0.1,0.1,0.1])
>>> hist, bin_edges = np.histogram(a,bins=10, weights=weights)
>>> print(bin_edges)
[  1.    1.9   2.8   3.7   4.6   5.5   6.4   7.3   8.2   9.1  10. ]
>>> hist, bin_edges = np.histogram(a,bins=10)
>>> print(bin_edges)
[  1.    1.9   2.8   3.7   4.6   5.5   6.4   7.3   8.2   9.1  10. ]

The bin_edges calculated with and without weights are the same. It seems the weights are ignored in the bin_edges calculation.

[nayyarv]

Hmm, when you pass in 10 bins, the bin edges are literally calculated by np.linspace(min(x), max(x), n_bins+1). This will never change depending on weights, since this does not change the range nor the number of bins.

What does change is the value (i.e. height) of each bin. That is affected by weights.

In [4]: >>> import numpy as np
   ...: >>> a=np.array([1,2,3,4,5,6,7,8,9,10])
   ...: >>> weights=np.array([0.8,0.1,0.1,0.1,0.1,0.1,0.1,0.1,0.1,0.1])
   ...: >>> hist, bin_edges = np.histogram(a,bins=10, weights=weights)
   ...: >>> print(hist)
[0.8 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1]
   ...: >>> hist, bin_edges = np.histogram(a,bins=10)
   ...: >>> print(hist)
[1 1 1 1 1 1 1 1 1 1]

And if you were to try the same with an automated method, you get a ValueError since the weights would change the estimate.

In [5]:  >>> hist, bin_edges = np.histogram(a,bins='auto', weights=weights)
TypeError                                 Traceback (most recent call last)
----> 1 hist, bin_edges = np.histogram(a,bins='auto', weights=weights)

/usr/local/lib/python2.7/site-packages/numpy/lib/function_base.pyc in histogram(a, bins, range, normed, weights, density)
    689                 "{!r} is not a valid estimator for `bins`".format(bin_name))
    690         if weights is not None:
--> 691             raise TypeError("Automated estimation of the number of "
    692                             "bins is not supported for weighted data")
    693         # Make a reference to `a`

TypeError: Automated estimation of the number of bins is not supported for weighted data

This is not a bug, this is a misunderstanding of how weights and bin edges interact.

[domagalski]

I agree. It can be closed.