Selective text annotations with plotnine¶

Sometimes you have a lot of datapoints on a graphic, but only want to annotate a few of them! That's where selective highlights and annotations come in handy.

In [15]:

import pandas as pd
from plotnine import *

df = pd.read_csv("countries.csv")
df.head()

import pandas as pd from plotnine import * df = pd.read_csv("countries.csv") df.head()

Out[15]:

	country	continent	gdp_per_capita	life_expectancy	population
0	Afghanistan	Asia	663	54.863	22856302
1	Albania	Europe	4195	74.200	3071856
2	Algeria	Africa	5098	68.963	30533827
3	Angola	Africa	2446	45.234	13926373
4	Antigua and Barbuda	N. America	12738	73.544	77656

In [16]:

df.shape

df.shape

Out[16]:

(188, 5)

Plot a subset of your data on top of existing data¶

If you want to annotate a subset of your data, it's usually best to create a separate dataframe just for the items you want to annotate.

In [17]:

annotations = df[df.country.isin(['Tonga', 'Angola', 'Iceland'])]
annotations

annotations = df[df.country.isin(['Tonga', 'Angola', 'Iceland'])] annotations

Out[17]:

	country	continent	gdp_per_capita	life_expectancy	population
3	Angola	Africa	2446	45.234	13926373
74	Iceland	Europe	31092	79.720	281210
169	Tonga	Oceania	4886	70.810	97935

Then, you first plot with your normal dataset, then add the annotations in on top. In this case we start by building a standard scatterplot, then pass the annotations dataframe to geom_point and geom_text to draw points and text just for that dataframe.

In [27]:

(
    ggplot(df)
    + aes(x='gdp_per_capita', y='life_expectancy', label='country')
    + geom_point(color='lightgrey')
    + geom_point(annotations, color='red')
    + geom_text(annotations, ha='left', nudge_x=1000)
)

( ggplot(df) + aes(x='gdp_per_capita', y='life_expectancy', label='country') + geom_point(color='lightgrey') + geom_point(annotations, color='red') + geom_text(annotations, ha='left', nudge_x=1000) )

Out[27]:

<ggplot: (310896580)>

By default, every geom_point or geom_text uses the df you sent to ggplot. If you give it a different dataset, though, it'll plot that different data!

In [ ]: