Visualization

There are a lot of different libraries for plotting and visualizing data, but the one we will focus on here is Matplotlib, which is a comprehensive library for creating static, animated, and interactive visualizations in Python.

More in-depth information is available at the official documentation and cheatsheets.

Setup

It is standard practice to import Matplotlib into the plt namespace:

import matplotlib.pyplot as plt

For the examples on this page we also need to import Polars and the MultipleLocator from Matplotlib:

import polars as pl
from matplotlib.ticker import MultipleLocator
from polars import col

We will be working with the patient data from the previous chapter:

patients = pl.read_csv("../../data/hosp/patients.csv.gz")

A Simple Figure

A figure in Matplotlib consists of a Figure object and one or more Axes objects. Each axis is an area available for plotting, contained within the figure.

Let’s start by exploring the age in the data as a histogram. We will create a figure with one axis where we can plot our histogram.

Code

fig, ax = plt.subplots()
ax.hist(patients["anchor_age"])

Output

Great! Now let’s increase the number of bins.

Code

fig, ax = plt.subplots()
ax.hist(patients["anchor_age"], bins=20)

Output

Making it Pretty

It looks ok, but it could definitely be prettier. Let’s change the color and make the bars stand out more with a slight border.

Code

fig, ax = plt.subplots()
ax.hist(patients["anchor_age"], bins=20, facecolor="#3b82f6", edgecolor="white")

Output

Much better! Continuing from the previous code, let’s set the limits and labels of our x and y axes. We will also tweak where the ticks are located on the y axis.

Code

ax.set_xlim(15, 95)
ax.set_ylim(0, 20)
ax.set_xlabel("Age in years")
ax.set_ylabel("Number of subjects")
ax.yaxis.set_major_locator(MultipleLocator(2))

Output

Finally, let’s add a grid underneath the plot to make it easier to see the scale of the bars.

Code

ax.grid(color="#f3f4f6")
ax.set_axisbelow(True)  # Puts the grid below the figure

Output

Now, isn’t that a beautiful figure!

Stratified by Sex

If we want to show multiple plots on the same axis, we can plot them one after the other using the same ax object. We better also add a legend for clarity.

Code

fig, ax = plt.subplots()

males = patients.filter(col("gender").eq("M"))["anchor_age"]
females = patients.filter(col("gender").eq("F"))["anchor_age"]

ax.hist(males, facecolor="#3b82f6", alpha=0.5, label="Male")
ax.hist(females, facecolor="#ef4444", alpha=0.5, label="Female")

ax.legend()

Output

If we instead want to split the stratification into separate axes/panels, we can tell the subplots function that we want 2 rows of axes and then plot the stratified data on the respective axis.

Code

fig, axs = plt.subplots(2, 1, constrained_layout=True)

axs[0].hist(males, facecolor="#3b82f6", alpha=0.5)
axs[0].set_title("Male")

axs[1].hist(females, facecolor="#ef4444", alpha=0.5)
axs[1].set_title("Female")

Output

Saving the Figure

Finally, we can save the figure for use in our manuscript (change the file path as needed). Many different file formats are supported (eg. png, svg, pdf, tiff).

fig.savefig("../../assets/img/visualization/figure-Pyw3e0.svg")

fig.savefig("../../assets/img/visualization/figure-Pyw3e0.png", dpi=300)

Different Types of Plots

Below are basic examples of different types of plots.

Line

Code

fig, ax = plt.subplots()

x = [0, 1, 2, 3]
y = [0.5, 12, 15, 3]

ax.plot(x, y)

Output

Bar

Code

fig, ax = plt.subplots()

ax.bar(*patients["gender"].value_counts())
ax.set_xlabel("Sex")
ax.set_ylabel("n")

Output

Histogram

Code

fig, ax = plt.subplots()
ax.hist(patients["anchor_age"])

Output

Box

Code

fig, ax = plt.subplots()

strat_by_age = patients.group_by("gender").agg("anchor_age")
ax.boxplot(strat_by_age["anchor_age"], tick_labels=strat_by_age["gender"].to_list())
ax.set_xlabel("Sex")
ax.set_ylabel("anchor_age")

Output

Scatter

Code

fig, ax = plt.subplots()

ax.scatter(patients["anchor_age"], patients["anchor_year"])
ax.set_xlabel("anchor_age")
ax.set_ylabel("anchor_year")

Output

Multiple Axes / Panels

We have already seen an example of a figure with two panels when we stratified the age histogram by sex above. Here we’ll look into more detail how to create multiple panels within one figure.

Code

fig, axs = plt.subplots(2, 2, constrained_layout=True)

axs[0][0].set_title("Histogram")
axs[0][0].hist(patients["anchor_age"])
axs[0][0].set_xlabel("Age in years")
axs[0][0].set_ylabel("Number of subjects")

axs[0][1].set_title("Bar")
axs[0][1].bar(*patients["gender"].value_counts())
axs[0][1].set_xlabel("Sex")
axs[0][1].set_ylabel("n")

strat_by_age = patients.group_by("gender").agg("anchor_age")
axs[1][0].set_title("Box")
axs[1][0].boxplot(
    strat_by_age["anchor_age"], tick_labels=strat_by_age["gender"].to_list()
)
axs[1][0].set_xlabel("Sex")
axs[1][0].set_ylabel("anchor_age")

axs[1][1].set_title("Scatter")
axs[1][1].scatter(patients["anchor_age"], patients["anchor_year"])
axs[1][1].set_xlabel("anchor_age")
axs[1][1].set_ylabel("anchor_year")

Output