Using matplotlib, how could one compare histograms by overlaying them and by showing their ratio plot?

Question

Two histograms can be compared by creating a plot that features both an overlay of the histograms (possibly normalised) and a ratio plot of the histograms. Here is such a plot:

How could a plot like this be made using matplotlib?

Answer 1

I don't see what the dots are, but here's a simple example of the ratios. The main trick is to reuse the bin values that hist returns.

import matplotlib.pyplot as plt
from numpy.random import normal

y = []
y.append(normal(2, 2, size=120))
y.append(normal(2, 2, size=120))

fig, (ax1, ax2) = plt.subplots(nrows=2)

ns, bins, patches = ax1.hist(y, normed=False,
                      histtype='stepfilled',
                      bins=8,
                      alpha=0.2,
                      label=['a','b']
                      )
ax1.legend()

ax2.bar(bins[:-1],     # this is what makes it comparable
        ns[0] / ns[1], # maybe check for div-by-zero!
        alpha=0.4)

ax1.set_ylabel('Data')
ax2.set_ylabel('Ratio (a/b)')