EDS 217 Cheatsheet

Customizing Plots with Matplotlib and Seaborn

This page covers the adjustments you make after a figure exists: sizing it, colouring it, labelling it, and making the labels readable. For the plotting commands themselves, see the matplotlib and seaborn cheatsheets.

Setup

Code 
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns

months = ['Jan','Feb','Mar','Apr','May','Jun','Jul','Aug','Sep','Oct','Nov','Dec']
temps = [-22.89, -20.70, -20.69, -11.76, -0.80, 8.59,
         11.22, 7.23, -0.11, -10.54, -18.34, -21.44]

Matplotlib Customization

Figure size

figsize is in inches, width then height. A wide, short figure suits a long time axis. A square one suits a scatter.

Code 
plt.figure(figsize=(10, 4))
plt.plot(months, temps)
plt.ylabel('Mean air temperature (deg C)')
plt.tight_layout()
plt.show()

Colour, line style and markers

Code 
plt.figure(figsize=(8, 4))
plt.plot(months, temps, color='firebrick', linestyle='--', linewidth=2,
         marker='o', markersize=5)
plt.ylabel('Mean air temperature (deg C)')
plt.tight_layout()
plt.show()

Common values: colours by name ('firebrick', 'steelblue') or single letter ('r', 'b', 'g', 'k'); line styles '-', '--', ':', '-.'; markers 'o', 's', '^', '.'.

Axis labels and title

Axis labels should give units. A title should say what the figure shows, not repeat the axis names.

Code 
plt.figure(figsize=(8, 4))
plt.bar(months, temps)
plt.xlabel('Month')
plt.ylabel('Mean air temperature (deg C)')
plt.title('Toolik Lake, 2008 to 2019')
plt.tight_layout()
plt.show()

Rotating tick labels

When category names collide, rotate them. ha='right' aligns each rotated label under its own tick rather than beside it.

Code 
long_labels = ['Great Smoky Mountains', 'Grand Canyon', 'Cuyahoga Valley',
               'Olympic', 'Rocky Mountain']
visitors = [6.069, 2.097, 2.093, 1.960, 1.765]

plt.figure(figsize=(7, 4))
plt.bar(long_labels, visitors)
plt.xticks(rotation=45, ha='right')
plt.ylabel('Mean annual visitors (millions)')
plt.tight_layout()
plt.show()

Before you rotate, ask whether a horizontal bar chart would read better. It usually does.

Code 
plt.figure(figsize=(7, 4))
plt.barh(long_labels, visitors)
plt.xlabel('Mean annual visitors (millions)')
plt.tight_layout()
plt.show()

Legends

Label each series as you draw it, then call legend() once.

Code 
early = [-24.1, -22.0, -21.5, -12.4, -1.2, 8.1, 11.6, 7.0, -1.3, -12.0, -19.5, -22.0]

plt.figure(figsize=(9, 4))
plt.plot(months, early, label='Early period')
plt.plot(months, temps, label='Late period')
plt.ylabel('Mean air temperature (deg C)')
plt.legend()
plt.tight_layout()
plt.show()

Grid lines

Code 
plt.figure(figsize=(8, 4))
plt.bar(months, temps)
plt.grid(axis='y', linestyle='--', alpha=0.7)
plt.ylabel('Mean air temperature (deg C)')
plt.tight_layout()
plt.show()

Fixing crowded layouts, and saving

Code 
plt.figure(figsize=(8, 4))
plt.bar(months, temps)
plt.ylabel('Mean air temperature (deg C)')
plt.tight_layout()     # keeps labels from being cut off
plt.show()

plt.savefig('toolik_monthly.png', dpi=300)   # call this before plt.show()

Seaborn Customization

Seaborn draws onto a matplotlib figure, so everything above still applies. Size the figure with plt.figure() first, then call seaborn, then label with plt.xlabel() and friends.

Code 
penguins = pd.read_csv('https://eds-217-essential-python.github.io/data/penguins.csv')
penguins = penguins.dropna()

Colouring by a category with hue=

Code 
plt.figure(figsize=(7, 5))
sns.scatterplot(data=penguins, x='bill_length_mm', y='bill_depth_mm', hue='species')
plt.xlabel('Bill length (mm)')
plt.ylabel('Bill depth (mm)')
plt.tight_layout()
plt.show()

Point size and transparency

s= sets the marker size and alpha= sets transparency. Lower the transparency when points overlap heavily.

Code 
plt.figure(figsize=(7, 5))
sns.scatterplot(data=penguins, x='flipper_length_mm', y='body_mass_g',
                hue='species', s=60, alpha=0.6)
plt.xlabel('Flipper length (mm)')
plt.ylabel('Body mass (g)')
plt.tight_layout()
plt.show()

Histogram bin count

bins= controls how finely the range is divided. Too few bins hides structure and too many turns the figure into noise. Try a few.

Code 
plt.figure(figsize=(7, 4))
sns.histplot(data=penguins, x='body_mass_g', bins=20)
plt.xlabel('Body mass (g)')
plt.tight_layout()
plt.show()

Ordering bars

A bar chart of unordered categories is easier to read when you sort it first.

Code 
mean_mass = penguins.groupby('species')['body_mass_g'].mean().sort_values()

plt.figure(figsize=(7, 3.5))
sns.barplot(x=mean_mass.values, y=mean_mass.index)
plt.xlabel('Mean body mass (g)')
plt.ylabel('Species')
plt.tight_layout()
plt.show()

Key Points

  1. Size the figure first, plot second, label third, tight_layout() last.
  2. Axis labels give units. A figure without them is not finished.
  3. Prefer horizontal bars to rotated labels when the categories are words.
  4. Sort a bar chart before you draw it.
  5. savefig() goes before show(), because showing a figure clears it.

Beyond EDS 217

Not used in this course, listed so you recognize it elsewhere. Not run here.

# Global styling: changes the look of every figure drawn afterwards
sns.set_theme(style='whitegrid')
sns.set_palette('deep')

# Named colour palettes for a hue variable
sns.scatterplot(data=df, x='a', y='b', hue='group', palette='Set2')

# Text and arrows placed at a data coordinate
plt.annotate('Trough', xy=(3, -1), xytext=(6, -0.5),
             arrowprops=dict(facecolor='black', shrink=0.05))

# Manual tick positions
plt.xticks([0, 2, 4, 6, 8, 10])
plt.ylim(-30, 15)

# Several panels in one figure
fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(12, 5))

See the seaborn cheatsheet for the seaborn functions this course does not use.