Categorical palettes

Improving Your Data Visualizations in Python

Nick Strayer

Instructor

three panels containing countries, cities, and birds as examples of categorical data

Limits in perception

Try and limit to 10 or fewer categories
Keep color-blindness in mind

sns.palplot(sns.color_palette('Set2', 11))

A bunch of colors that are hard to tell apart due to how many there are

# Assign a new column to dataframe the desired combos
pollution['interesting cities'] = [x if x in ['Long Beach', 'Cincinnati'] 
                                   else 'other' for x in pollution['city'] ]

sns.scatterplot(x="NO2", y="SO2", hue = 'interesting cities', palette='Set2',
                data=pollution.query('year == 2014 & month == 12'))

scatter plot with Long Beach and Cincinnati given distinct point colors and all other cities lumped together into an other color

colorbrewer_palettes = ['Set1',   'Set2',    'Set3',    'Accent', 
                        'Paired', 'Pastel1', 'Pastel2', 'Dark2']

for pal in colorbrewer_palettes: 
    sns.palplot(pal=sns.color_palette(pal))
    plt.title(pal, loc = 'left')

A series of available categorical color palettes

Ordinal data (a)

Has order between classes
A set number of distinct classes

Diagram showing four quartiles

Ordinal data (b)

Has order between classes
A set number of distinct classes

Diagram showing the seven days of the week

Ordinal data (c)

Has order between classes
A set number of distinct classes

Diagram showing scale of happy to sad using emoji faces

colorbrewer_palettes = ['Reds', 'Blues', 'YlOrBr', 'PuBuGn', 'GnBu', 'Greys']

for i, pal in enumerate(colorbrewer_palettes): 
    sns.palplot(pal=sns.color_palette(pal, n_colors=i+4))

A series of available ordinal palettes

# Make a tertials column using qcut()
pollution['NO2 Tertial'] = pd.qcut(pollution['NO2'], 3, labels = False)

# Plot colored by the computer tertials 
sns.scatterplot(x="CO", y="SO2", hue='NO2 Tertial', palette="OrRd",
                data=pollution.query("city == 'Long Beach' & year == 2014"))

Scatterplot of CO and SO2 values that encodes the tertials of NO2 in a red ordinal color scale

Let's color some categories

Improving Your Data Visualizations in Python