Back to Article
Basic Density Plot
Download Notebook
In [1]:
import pandas as pd
import numpy as np

from plotnine import (
    ggplot,
    aes,
    after_stat,
    geom_density,
    geom_histogram,
    geom_vline,
    geom_rect,
    labs,
    annotate,
    theme_tufte,
)
from plotnine.data import mpg
In [2]:
mpg.head()
manufacturer model displ year cyl trans drv cty hwy fl class
0 audi a4 1.8 1999 4 auto(l5) f 18 29 p compact
1 audi a4 1.8 1999 4 manual(m5) f 21 29 p compact
2 audi a4 2.0 2008 4 manual(m6) f 20 31 p compact
3 audi a4 2.0 2008 4 auto(av) f 21 30 p compact
4 audi a4 2.8 1999 6 auto(l5) f 16 26 p compact

The defaults are not exactly beautiful, but still quite clear.

Basic Density Plot

In [3]:
# Gallery, distributions
(
    ggplot(mpg, aes(x="cty"))
    + geom_density()
)

Plotting multiple groups is straightforward, but as each group is plotted as an independent PDF summing to 1, the relative size of each group will be normalized.

Density Plot with Groups

In [4]:
# Gallery, distributions
(
    ggplot(mpg, aes(x="cty", color="drv", fill="drv"))
    + geom_density(alpha=0.1)
)

To plot multiple groups and scale them by their relative size, you can map the y aesthetic to 'count' (calculated by stat_density).

In [5]:
(
    ggplot(mpg, aes(x="cty", color="drv", fill="drv"))
    + geom_density(aes(y=after_stat("count")), alpha=0.1)
)

Density Plots + Histograms

To overlay a histogram onto the density, the y aesthetic of the density should be mapped to the 'count' scaled by the binwidth of the histograms.

Why?

The count calculated by stat_density is \(count = density * n\) where n is the number of points . The density curves have an area of 1 and have no information about the absolute frequency of the values along curve; only the relative frequencies. The count curve reveals the absolute frequencies. The scale of this count corresponds to the count calculated by the stat_bin for the histogram when the bins are 1 unit wide i.e. binwidth=1. The count * binwidth curve matches the scale of counts for the histogram for a give binwidth.

In [6]:
# Gallery, distributions

binwidth = 2  # The same for geom_density and geom_histogram

(
    ggplot(mpg, aes(x="cty", color="drv", fill="drv"))
    + geom_density(aes(y=after_stat("count*binwidth")), alpha=0.1)
    + geom_histogram(
        aes(fill="drv", y=after_stat("count")),
        binwidth=binwidth,
        color="none",
        alpha=0.5,
    )
    # It is the histogram that gives us the meaningful y axis label
    # i.e. 'count' and not 'count*2'
    + labs(y="count")
)

Shaded Range Under a Density Plot

Extending geom_density to create an effect of a shaded range

Create some data and plot the density

In [7]:
n = 101
df = pd.DataFrame({"x": np.arange(n)})

(
    ggplot(df, aes("x"))
    + geom_density()
)

Suppose we want to mark a region as special e.g. (40, 60), we can use vertical lines to annotate it.

In [8]:
region = (40, 60)

(
    ggplot(df, aes("x"))
    + geom_density()
    + annotate(geom_vline, xintercept=region)  # new line
)

To make it standout more we can highlight. To do that, the first thought is to use a rectangle.

In [9]:
region = (40, 60)

(
    ggplot(df, aes("x"))
    + geom_density()
    + annotate(
        geom_rect, xmin=region[0], xmax=region[1], ymin=0, ymax=float("inf"), alpha=0.5
    )  # new annotation layer
    + annotate(geom_vline, xintercept=region)
)

Since y upper-bound varies along the curve, a rectangular highlight has to stretch up to the top of the panel.

To hightlight only within the density curve, we have to use a second density curve. We need to calculate the density as normal, but just before the curve & region are plotted, we should keep only the region we want.

We create our own geom_density_highlight and override the setup_data method. First, we override but do nothing, we only inspect the data to see what we have to work with.

In [10]:
# new class
class geom_density_highlight(geom_density):
    def setup_data(self, data):
        data = super().setup_data(data)
        print(data)
        return data


region = (40, 60)

(
    ggplot(df, aes("x"))
    + geom_density()
    + geom_density_highlight(fill="black", alpha=0.5)  # new line
    + annotate(geom_vline, xintercept=region)
)
      PANEL     count   density  group    n    scaled           x         y  \
0         1  0.519038  0.005139     -1  101  0.519039    0.000000  0.005139   
1         1  0.522757  0.005176     -1  101  0.522758    0.097752  0.005176   
2         1  0.526473  0.005213     -1  101  0.526474    0.195503  0.005213   
3         1  0.530187  0.005249     -1  101  0.530188    0.293255  0.005249   
4         1  0.533899  0.005286     -1  101  0.533900    0.391007  0.005286   
...     ...       ...       ...    ...  ...       ...         ...       ...   
1019      1  0.533899  0.005286     -1  101  0.533900   99.608993  0.005286   
1020      1  0.530187  0.005249     -1  101  0.530188   99.706745  0.005249   
1021      1  0.526473  0.005213     -1  101  0.526474   99.804497  0.005213   
1022      1  0.522757  0.005176     -1  101  0.522758   99.902248  0.005176   
1023      1  0.519038  0.005139     -1  101  0.519039  100.000000  0.005139   

      ymin      ymax  
0        0  0.005139  
1        0  0.005176  
2        0  0.005213  
3        0  0.005249  
4        0  0.005286  
...    ...       ...  
1019     0  0.005286  
1020     0  0.005249  
1021     0  0.005213  
1022     0  0.005176  
1023     0  0.005139  

[1024 rows x 10 columns]

The highlight has filled the whole region, but the printed data suggests that we can limit the rows to those where x column is within our region.

In [11]:
class geom_density_highlight(geom_density):
    # new method
    def __init__(self, *args, region=(-np.inf, np.inf), **kwargs):
        super().__init__(*args, **kwargs)
        self.region = region

    def setup_data(self, data):
        data = super().setup_data(data)
        s = f"{self.region[0]} <= x <= {self.region[1]}"  # new line
        data = data.query(s).reset_index(drop=True)  # new line
        return data


region = (40, 60)

(
    ggplot(df, aes("x"))
    + geom_density()
    + geom_density_highlight(region=region, fill="black", alpha=0.5)  # modified line
    + annotate(geom_vline, xintercept=region)
)

That is it, but we can make it look better.

In [12]:
# Gallery, distributions

class geom_density_highlight(geom_density):
    def __init__(self, *args, region=(-np.inf, np.inf), **kwargs):
        super().__init__(*args, **kwargs)
        self.region = region

    def setup_data(self, data):
        data = super().setup_data(data)
        s = f"{self.region[0]} <= x <= {self.region[1]}"
        data = data.query(s).reset_index(drop=True)
        return data


region = (40, 60)
teal = "#029386"


(
    ggplot(df, aes("x"))
    + geom_density_highlight(region=region, fill=teal + "88", color="none")
    + geom_density(fill=teal + "44", color=teal, size=0.7)
    + annotate(geom_vline, xintercept=region, color=teal, size=0.7)
    + theme_tufte()
)

This example was motivated by a question from github user Rishika-Ravindran.