lesson 217 Visualization & EDA · advanced

Charts from DataFrames

pandas can draw a DataFrame or Series directly with .plot, using matplotlib underneath: one call turns a pivot table into grouped bars. Pass ax= to choose where it draws.

You'll be able to

Draw a Series or DataFrame with .plot(kind=...) or .plot.bar() and .plot.barh(), onto a chosen ax
Explain how a DataFrame's index and columns become the x-axis and the legend
Lay out several charts with plt.subplots(nrows, ncols) and draw one on each Axes

The idea

After a groupby or pivot_table the numbers are already in a DataFrame or Series, labelled. .plot draws them with matplotlib in one call: the index becomes the x-axis, and each column becomes one series (one colour of bar, one line), named in a legend. kind= picks the chart ("bar", "barh", "line", "hist", "scatter"), or use the method form .plot.bar(). It returns the Axes it drew on; pass ax= to draw on one you made.

You may meet seaborn, a separate library built on matplotlib for statistical charts. It isn't installed in this course's browser Python, so the course draws the same kinds of chart with pandas and matplotlib.

Run each example to see what it prints. Running examples doesn't count toward progress.

example-1.py · A pivot table as grouped bars

import matplotlib.pyplot as plt
import pandas as pd

orders = pd.DataFrame({
    "ship_state": ["TX", "CA", "NY", "TX", "NY", "TX"],
    "channel": ["web", "web", "web", "web", "phone", "app"],
    "total": [35.84, 91.47, 55.72, 246.6, 54.55, 39.15],
})
table = orders.pivot_table(index="ship_state", columns="channel", values="total", aggfunc="sum", fill_value=0)
fig, ax = plt.subplots()
drawn = table.plot(kind="bar", ax=ax, title="Revenue by state and channel")
print(drawn is ax, len(ax.patches))
print([t.get_text() for t in ax.get_legend().get_texts()])
print([t.get_text() for t in ax.get_xticklabels()], ax.get_xlabel())
plt.show()

example-2.py · A scatter plot with only one column

import pandas as pd

films = pd.DataFrame({"budget_musd": [41.1, 43.1, 9.8], "revenue_musd": [105.8, 207.4, 10.7]})
films.plot(kind="scatter", x="budget_musd")

plt.subplots(nrows, ncols) makes a grid: axes is then an array of Axes, one per panel (axes[0], axes[1] for one row). Draw each chart on its own Axes with ax=. A shared y-axis (sharey=True) makes panels comparable.

example-3.py · Two panels: runtimes and films per decade

import matplotlib.pyplot as plt
import pandas as pd

films = pd.DataFrame({
    "title": ["The Country of Glass", "The Harbor on Fire", "The Island at Dawn", "Quiet Orchard", "Silent Kingdom", "Last Tide"],
    "release_year": [2015, 1996, 2014, 2019, 1987, 2003],
    "runtime_min": [87, 133, 117, 79, 107, 69],
})
fig, axes = plt.subplots(1, 2)
films["runtime_min"].plot.hist(ax=axes[0], bins=[60, 90, 120, 150], title="Runtime (min)")
decade = films["release_year"] // 10 * 10
decade.value_counts().sort_index().plot.bar(ax=axes[1], title="Films per decade")
print(len(fig.axes), [a.get_title() for a in fig.axes])
print([float(p.get_height()) for p in axes[0].patches])
print([t.get_text() for t in axes[1].get_xticklabels()], [int(p.get_height()) for p in axes[1].patches])
plt.show()

When to use it: quick charts of tables you already grouped, and anything where index and columns are the chart's labels. Switch to ax methods for fine control. Mistakes to avoid: plotting an unsorted value_counts when the categories have a natural order; forgetting ax= (pandas then draws on whichever Axes is current, usually the last one created, not the panel you meant); and plotting a wide table when you meant one column (df["col"].plot).

Review

df.plot(kind=..., ax=...) and series.plot.bar() draw with matplotlib and return the Axes.
Index → x-axis; columns → series and legend.
plt.subplots(nrows, ncols) gives a grid of Axes; draw each panel with ax=axes[i].
Next: which chart answers which question.

Practise this