Data Visualization
The same five numbers, two ways
In the pie the slices look equal. In the bars the decline from A to E is obvious.
Figure and Axes give the canvas
pilot = pd.DataFrame({
"sleep": [6.5, 7.5, 5.5, 8.0, 6.0],
"stress": [3.2, 2.1, 4.5, 1.8, 3.9]
})
fig, ax = plt.subplots(figsize=(5, 3.5))
ax.set_xlabel("Sleep (hours)")
ax.set_ylabel("Stress (1 to 5)")
plt.show()The axes exist, but nothing is drawn yet.
Seaborn draws the data onto the axes
fig, ax = plt.subplots(figsize=(5, 3.5))
sns.scatterplot(
data=pilot,
x="sleep",
y="stress",
s=80,
color="#1d588c",
ax=ax
)
plt.show()One point per row. Passing ax=ax connects Seaborn directly to the Matplotlib canvas.
Histograms show a distribution
fig, ax = plt.subplots(figsize=(5, 3.5))
sns.histplot(
data=first_sem,
x="sleep_hours",
binwidth=0.5,
color="steelblue",
edgecolor="white",
ax=ax
)Bell-shaped, centred a little above 6 hours.
r round(100 * (first_sem['sleep_hours'] < 7).mean())% of students sleep less than the recommended 7.
The binwidth changes the picture
Too narrow is noisy, too wide hides the shape. Try a few widths.
Count charts count categories
fig, ax = plt.subplots(figsize=(5, 3.5))
sns.countplot(
data=students,
y="faculty",
color="#1d588c",
ax=ax
)sns.countplot() counts rows automatically.
For precomputed values such as averages, use sns.barplot().
Box plots compare groups
fig, ax = plt.subplots(figsize=(5, 3.5))
sns.boxplot(
data=first_sem,
x="study_mode",
y="wellbeing",
color="#eaf0f7",
ax=ax
)Part-time students sit a little lower, but the boxes overlap considerably.
Scatter plots show relationships
fig, ax = plt.subplots(figsize=(5, 3.5))
sns.regplot(
data=first_sem,
x="sleep_hours",
y="gpa",
scatter_kws={"alpha": 0.35, "color": "#1d588c"},
line_kws={"color": "#b4762e"},
ax=ax
)alpha makes points transparent against overplotting.
The line rises gently, but points scatter widely.
Line charts show change over time
fig, ax = plt.subplots(figsize=(5, 3.5))
sns.lineplot(
data=wellbeing_by_sem,
x="semester",
y="wellbeing",
hue="workshop",
marker="o",
palette=["#1d588c", "#b4762e"],
ax=ax
)Shows the longitudinal trajectory across all four study periods.
A common visualization mistake
# Unnecessary mapping for one constant
sns.barplot(
data=faculty_means,
x="faculty",
y="wellbeing",
hue="faculty" # redundant!
)Using hue for categorical variables already labeled on the axis creates visual clutter and a useless legend.
Annotations give immediate reference
fig, ax = plt.subplots(figsize=(5, 3.5))
sns.histplot(data=first_sem, x="sleep_hours", binwidth=0.5, ax=ax)
ax.axvline(7, color="darkred", linestyle="--")
ax.text(7.1, 75, "Recommended: 7h", color="darkred")Guides the eye directly to the clinical or theoretical threshold.
A misleading graph: truncated baseline
Education appears several times taller than Social Sciences, yet actual averages differ by only 3 points out of 100!
The same data, redesigned honestly
Differences between faculties are negligible compared to the substantial variation among individual students.
Facets: small multiples
g = sns.FacetGrid(
first_sem,
col="programme",
col_wrap=3,
height=2.8
)
g.map_dataframe(
sns.scatterplot,
x="sleep_hours",
y="gpa",
alpha=0.4
)Identical shared axes enable rapid, accurate visual comparisons.
Density plots compare whole distributions
sem2 = (
semesters[semesters["semester"] == 2]
.merge(students, on="student_id")
)
fig, ax = plt.subplots(figsize=(5, 3.5))
sns.kdeplot(
data=sem2,
x="wellbeing",
hue="workshop",
fill=True,
alpha=0.3,
palette=[navy, copper],
ax=ax
)Shows that the entire distribution shifted upward, not just a handful of students.
Correlation heatmaps
A diverging palette centered at 0 with overlaid numeric coefficients prevents misreading subtle color tints.