Skip to content

Session 13 — Matplotlib: Figures, Axes, Customization

Week 7 · Session 13 · Module A · 90 min · CLO-1

1. Learning objectives

By the end of this session, students can: - Explain why visualization is a core analysis tool, not decoration. - Create line, bar, scatter, and histogram plots with plt/ax. - Use the figure–axes model to control multiple plots in one figure. - Customize labels, titles, colors, grid, legend, and limits. - Save publication-quality figures with savefig and dpi. - Choose the right chart for the question (relationship → scatter; distribution → histogram; categories → bar).

2. Key concepts

  • A chart answers a question — pick the chart for the question, not the prettiest one.
  • Figure vs. axes: figure = canvas; axes = one plot area (a figure holds 1..n axes).
  • The pyplot interface (plt.plot(...)) is convenient; the object interface (fig, ax = plt.subplots()) is explicit and scales to subplots.
  • Three workhorse charts: line (trend over time), scatter (relationship between two numerics), bar (category values), histogram (distribution of one numeric).
  • %matplotlib inline embeds figures in the notebook.
  • Label everything: title, x-label, y-label, legend — an unlabeled chart is noise.

3. Detailed lecture notes

Why visualize? Numbers hide patterns; pictures reveal them. Anscombe's quartet is the canonical demo: four datasets with identical mean, variance, correlation, and regression line — but wildly different scatter plots. If you only summarize, you miss the point; if you plot, you see it. Visualization is therefore part of exploration (finding problems: outliers, clusters, missing patterns) and communication (the final lifecycle stage).

The mental model: figure and axes. A figure is the canvas; axes are the plot areas on it. plt.plot(x, y) creates a figure with one axes implicitly; the explicit form gives control: 

fig, ax = plt.subplots()        # one canvas, one plot
ax.plot(x, y)
fig, axes = plt.subplots(1, 2)  # one canvas, two plots side by side
Every customization is a method on ax: set_title, set_xlabel, set_ylabel, legend, grid, set_xlim, set_ylim. Colors via color=, style via linestyle=, marker=. Teach the explicit form early — it makes subplots trivial later instead of a mystery.

Choosing the chart. The single most useful decision framework: - Trend over time (ordered x) → line plot. - Two numeric variables, relationship? → scatter plot. - Categories vs. numeric values → bar plot. - Distribution of one numeric variable → histogram (ax.hist). - Part-to-whole → pie (avoid — bars are easier to read; say so briefly). Show each on the same dataset (tips): line isn't great for tips (no time axis — use flights), scatter of total_bill vs tip reveals the relationship, bar of mean tip by day, histogram of tip.

Taste and honesty. A few principles: label every axis with units; never truncate a bar chart's y-axis to exaggerate; use color meaningfully (not rainbow); 3D charts are usually worse than 2D. Data visualization ethics = don't mislead. Mention this connects to responsible AI/communication (Session 30).

Saving and sharing. fig.savefig("charts/tips-scatter.png", dpi=150, bbox_inches="tight") — commit charts next to notebooks when a report needs them. dpi controls resolution; bbox_inches="tight" trims whitespace.

4. Important terminology

  • Figure — the canvas that holds axes.
  • Axes — one plot area (labels, ticks, data); a figure can contain many.
  • pyplot (plt) — the stateful convenience interface.
  • Object interface — fig, ax = plt.subplots(); explicit and preferred for control.
  • Series/DataFrame .plot() — Pandas' shortcut that calls matplotlib for you.
  • Histogram — bars over binned numeric ranges; shows distribution.
  • Scatter plot — points at (x, y) pairs; shows relationships/outliers.
  • Legend — key mapping colors/markers to series.
  • savefig — export the figure to PNG/PDF/SVG.
  • Anscombe's quartet — the four datasets proving summary statistics can lie without plots.

5. Python examples

import matplotlib.pyplot as plt
import seaborn as sns
import numpy as np

%matplotlib inline

# --- The four workhorse charts, one axes each ---
fig, axes = plt.subplots(2, 2, figsize=(10, 7))

# 1. Line: trend over time
flights = sns.load_dataset("flights")
monthly = flights.groupby("month")["passengers"].sum()
axes[0, 0].plot(monthly.index, monthly.values, marker="o")
axes[0, 0].set_title("Line: passengers by month")
axes[0, 0].set_xlabel("Month"); axes[0, 0].set_ylabel("Passengers")

# 2. Scatter: relationship
tips = sns.load_dataset("tips")
axes[0, 1].scatter(tips["total_bill"], tips["tip"], alpha=0.6)
axes[0, 1].set_title("Scatter: bill vs tip")
axes[0, 1].set_xlabel("Total bill ($)"); axes[0, 1].set_ylabel("Tip ($)")

# 3. Bar: categories
by_day = tips.groupby("day")["tip"].mean()
axes[1, 0].bar(by_day.index, by_day.values, color="teal")
axes[1, 0].set_title("Bar: mean tip by day")

# 4. Histogram: distribution
axes[1, 1].hist(tips["tip"], bins=20, edgecolor="white")
axes[1, 1].set_title("Histogram: distribution of tips")

fig.tight_layout()
fig.savefig("charts/tips-overview.png", dpi=150, bbox_inches="tight")

6. Beginner example

import matplotlib.pyplot as plt

x = [1, 2, 3, 4, 5]
y = [2, 4, 1, 8, 7]

plt.plot(x, y)
plt.title("My first chart")
plt.xlabel("X"); plt.ylabel("Y")
plt.show()

Five lines — and every later chart is this same shape with more options.

7. Practical Data Science example

import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd

%matplotlib inline
tips = sns.load_dataset("tips")

# Question: "Do bigger parties tip a higher percentage?"
# (percent tip is the fair comparison across bill sizes)
tips["tip_pct"] = tips["tip"] / tips["total_bill"] * 100

fig, ax = plt.subplots(figsize=(8, 5))
ax.scatter(tips["size"], tips["tip_pct"], alpha=0.6)
ax.set_title("Tip percentage vs. party size")
ax.set_xlabel("Party size (people)")
ax.set_ylabel("Tip percentage (%)")
ax.grid(alpha=0.3)

# Annotation: the honest observation
ax.axhline(tips["tip_pct"].mean(), color="red", linestyle="--",
           label=f'mean = {tips["tip_pct"].mean():.1f}%')
ax.legend()
fig.savefig("charts/tip-pct-by-size.png", dpi=150, bbox_inches="tight")

8. In-class activity (50 min)

In notebooks/week-07/session-13-matplotlib.ipynb:

  1. Four charts (20 min): reproduce the four-workhorse-chart example on tips — but choose your own x/y combinations and write a one-line takeaway under each.
  2. Customization (15 min): pick one chart and add: title, axis labels with units, grid, legend, and a color that isn't the default; then a second subplot with a different chart for the same question.
  3. Chart choice (10 min): given three questions (trend, relationship, distribution), predict the chart type; then plot one of them.
  4. Save & check (5 min): savefig to charts/; confirm the file exists and opens.

9. Lab exercise

Lab 13 (../labs/lab-13-matplotlib.md) — due before Session 14: build the 2×2 dashboard figure on tips, label every axis, and save it at 150 dpi. (Quiz 2 runs later: Session 29, covering Weeks 8–14.)

10. Common mistakes

  • Plotting without labels → a chart no one (including you, later) can read.
  • plt.show() before savefig in some setups → empty saved file (order matters in stateful mode).
  • Choosing a line plot for categorical data → meaningless zig-zags.
  • Forgetting %matplotlib inline → charts in a separate window or missing from the notebook.
  • Ignoring figsize → tiny, unreadable defaults in notebooks.
  • Overplotting: 10,000 points with no alpha → a solid blob hiding the pattern.
  • Pie charts for comparisons — bars are almost always clearer.

11. Short assessment questions

  1. What is the difference between a figure and an axes?
  2. Which chart type for: (a) temperature over a week; (b) relationship between height and weight; (c) most common grade in a class?
  3. What does bins control in a histogram?
  4. Why is the explicit fig, ax = plt.subplots() form better than plt.plot alone?
  5. How do you save a figure at high resolution with trimmed whitespace?
  6. True/False: two datasets with identical summary statistics always look similar when plotted. (False — Anscombe's quartet.)

12. CLO mapping

CLO-1: visualization is the "visualize" verb of CLO-1 and half of EDA. The figure–axes skills here feed directly into the Seaborn session (14), the EDA case study (15), and the final project's exploration section.

13. Suggested homework

  • Finish Lab 4 and commit (due Session 14).
  • Practice: recreate any chart from your Assignment 1 data with custom labels and saved output.
  • Read: Matplotlib "Pyplot tutorial" (matplotlib.org) — the first half only.
  • Preview: sns.scatterplot(data=tips, x="total_bill", y="tip", hue="day") — Session 14 shows why Seaborn makes this one line.