Drawing histograms with matplotlib

The notebook source code is here. Open this notebook at Google Colab.

Some preparation

import matplotlib
import matplotlib.pyplot as plt
import numpy as np

# use a smaller fig size
fig_size = plt.rcParams['figure.figsize']
print(f"Original fig size: {fig_size}")
plt.rcParams['figure.figsize'] = fig_size = [x/2 for x in fig_size]
print(f"New fig size: {fig_size}")

Original fig size: [6.4, 4.8]
New fig size: [3.2, 2.4]

A simple histogram

A histogram counts the number of occurrences of data within different ranges.

Use plt.hist(data) to draw a very basic histogram.

data = []
for i in range(10):
    data += [i] * (10 - i)

plt.hist(data)
plt.show()

Adding labels

Pass the return value of plt.hist() to plt.bar_label() to add value count to bars.

n, edges, bars = plt.hist(data)
plt.bar_label(bars, fontsize=10, padding=2)
plt.show()

The values are a bit too high, enlarge ylim. (Or you can hide the borders)

n, edges, bars = plt.hist(data)
plt.bar_label(bars, fontsize=10, padding=2)
plt.ylim(0, 11.5)
plt.show()

Prettify bars

Reduce bar width, so as to separate them with blanks. rwidth is a float number that sets the relative width of the bars as a fraction of the bin width.

n, edges, bars = plt.hist(data, rwidth=0.8)
plt.bar_label(bars, fontsize=10, padding=2)
plt.ylim(0, 11.5)
plt.show()

What if we want some bars to have a different color? Again, use the bars returned by plt.hist().

n, edges, bars = plt.hist(data, rwidth=0.8)
plt.bar_label(bars, fontsize=10, padding=2)
plt.ylim(0, 11.5)
bars[0].set_facecolor('r')
plt.show()

A different orientation

Sometimes, the data can be so large that value counts overlap with each other (or other bars).

data = []
for i in range(10):
    data += [i] * (10 - i) * 10000
data = np.asarray(data)

n, edges, bars = plt.hist(data, rwidth=0.8)
plt.bar_label(bars, fontsize=10, padding=2)
plt.ylim(0, 11.5 * 10000)
bars[0].set_facecolor('r')
plt.show()

This calls for a horizontal histogram. Pass orientation="horizontal" to plt.hist().

n, edges, bars = plt.hist(data, rwidth=0.8, orientation="horizontal")
plt.bar_label(bars, fontsize=10, padding=2)
plt.xlim(0, 130000)
bars[0].set_facecolor('r')
plt.show()

I want to reverse the y-axis. Obtain axis with fig, ax = plt.subplots() and use ax.invert_yaxis() to do the trick.

fig, ax = plt.subplots()
n, edges, bars = ax.hist(data, rwidth=0.8, orientation="horizontal")
ax.invert_yaxis()
plt.bar_label(bars, fontsize=10, padding=2)
plt.xlim(0, 130000)
bars[0].set_facecolor('r')
plt.show()

Change the bins

From the doc

bins: int or sequence or str, default: rcParams["hist.bins"] (default: 10)

• If bins is an integer, it defines the number of equal-width bins in the range.

• If bins is a sequence, it defines the bin edges, including the left edge of the first bin and the right edge of the last bin; in this case, bins may be unequally spaced. All but the last (righthand-most) bin is half-open. In other words, if bins is:

  [1, 2, 3, 4]
  

  then the first bin is [1, 2) (including 1, but excluding 2) and the second [2, 3). The last bin, however, is [3, 4], which includes 4.

• If bins is a string, it is one of the binning strategies supported by numpy.histogram_bin_edges: 'auto', 'fd', 'doane', 'scott', 'stone', 'rice', 'sturges', or 'sqrt'.

Bins as an integer

When bins is an integer, it controls the number of bins. By default, there are 10 bins. So if data is in the range [0, 20), the first bin counts the range [0, 2), the second [2, 4), the third [4, 6), and so forth.

data = []
for i in range(20):
    data += [i] * (20 - i)

plt.hist(data)
plt.show()

Changing bin number to 5 makes each bin count 4 numbers. The first bin now counts [0, 4).

plt.hist(data, bins=5)
plt.show()

Bins as a sequence

Now, the length of bins controls the bin number.

I am interested in the count of the smallest number (0), the middle ones (9, 10), and the largest one (19). I want to count them separately from the rest of the range.

This makes the range unevenly spaced: [0, 1), [1, 5), [5, 9), [9, 11), [11, 15), [15, 19), [19, 19].

data = []
for i in range(20):
    data += [i] * (20 - i)
# add some 19s so the graph looks better
data += [19]*20

bins = [0, 1, 5, 9, 11, 15, 19, 19]
n, edges, bars = plt.hist(data, bins=bins, rwidth=0.8)
plt.bar_label(bars, fontsize=10, padding=2)
for i in [0, 3, 6]:
    bars[i].set_facecolor('r')
plt.show()

Notice that last bin? It's become invisible!

That's because the last range ([19, 19]) has zero width! To fix this, make the range [19, 20], with a width of one.

bins = [0, 1, 5, 9, 11, 15, 19, 20]
n, edges, bars = plt.hist(data, bins=bins, rwidth=0.8)
plt.bar_label(bars, fontsize=10, padding=2)
plt.ylim(0, 80)
for i in [0, 3, 6]:
    bars[i].set_facecolor('r')
plt.show()