Question

I need to plot a bar graph with asymmetrical error bars…

The documentation of the matplotlib.pyplot.bar function says:

Detail: xerr and yerr are passed directly to
errorbar(), so they can also have shape 2xN for independent
specification of lower and upper errors.

But, I can not give an 2xN array to the yerr…

import numpy as np
import matplotlib.pyplot as plt

plt.bar(xrange(5), [2,5,3,4,7], yerr=[[1,4,2,3,6],[4,10,6,8,14]]) #DO NOT work!

And show me the following error:

Traceback (most recent call last):
  File "bar_stacked.py", line 9, in <module>
    plt.bar(xrange(5), [2,5,3,4,7], yerr=[[1,4,2,3,6],[4,10,6,8,14]])
  File "/usr/lib/pymodules/python2.7/matplotlib/pyplot.py", line 1742, in bar
    ret = ax.bar(left, height, width, bottom, color, edgecolor, linewidth, yerr, xerr, ecolor, capsize, align, orientation, log, **kwargs)
  File "/usr/lib/pymodules/python2.7/matplotlib/axes.py", line 4253, in bar
    "incompatible sizes: bar() argument 'yerr' must be len(%s) or scalar" % nbars)
ValueError: incompatible sizes: bar() argument 'yerr' must be len(5) or scalar

But, instead this function:

import numpy as np
import matplotlib.pyplot as plt

plt.errorbar(xrange(5), [2,5,3,4,7], yerr=[[1,4,2,3,6],[4,10,6,8,14]])

Works fine.

Does the matplotlib.pyplot.bar no longer support the 2xN arrays for yerr?
If the answere is yes… How can I plot a bar graph with asymmetrical error bars?

Thanks for your time!

Answer 1

Which version of matplotlib are you using?

With the 1.1.1 (latest stable) version your code work flawlessly.