Skip to content

Box Plot Keywords

The ax.boxplot() method accepts numerous keyword arguments to control appearance and behavior.

Mental Model

Box plot keywords fall into three groups: data interpretation (vert, whis, showfliers), labeling (labels, positions), and styling (patch_artist, boxprops, medianprops). The most impactful is patch_artist=True, which switches boxes from outlines to filled patches, unlocking color customization.

Most Important Parameters

If you learn only four keywords, make it these:

  1. whis — controls outlier definition (default 1.5×IQR; set to [5, 95] for percentile-based whiskers)
  2. showfliers — toggle outlier dots on/off (hide them to reduce noise in dense data)
  3. notch — adds a confidence interval notch around the median (non-overlapping notches suggest different medians)
  4. patch_artist — switches boxes to filled patches (required for color customization)

Labels

The labels keyword assigns names to each box on the x-axis.

1. Basic Labels

```python import matplotlib.pyplot as plt import numpy as np

np.random.seed(42) data_a = np.random.normal(100, 10, 200) data_b = np.random.normal(90, 20, 200) data_c = np.random.normal(110, 15, 200)

fig, ax = plt.subplots() ax.boxplot([data_a, data_b, data_c], labels=['Group A', 'Group B', 'Group C']) plt.show() ```

2. LaTeX Labels

python ax.boxplot([data_a, data_b, data_c], labels=[r'$10^4$', r'$5 \cdot 10^4$', r'$10^5$'])

3. Numeric Labels

python ax.boxplot([data_a, data_b, data_c], labels=['N=100', 'N=500', 'N=1000'])

Widths

The widths keyword controls the width of each box.

1. Uniform Width

python ax.boxplot([data_a, data_b, data_c], widths=0.5) # All boxes same width

2. Variable Widths

python ax.boxplot([data_a, data_b, data_c], widths=[0.2, 0.4, 0.6])

3. Proportional to Sample Size

python sizes = [100, 500, 1000] widths = [np.sqrt(s) / np.sqrt(max(sizes)) * 0.8 for s in sizes] ax.boxplot([data_a, data_b, data_c], widths=widths)

Orientation

The vert keyword controls vertical or horizontal orientation.

1. Vertical (Default)

python ax.boxplot(data, vert=True)

2. Horizontal

python fig, ax = plt.subplots() ax.boxplot([data_a, data_b, data_c], vert=False) ax.set_yticklabels(['A', 'B', 'C']) ax.set_xlabel('Value') plt.show()

3. Use Case

Horizontal orientation works well when labels are long or when comparing many groups.

Notch

The notch keyword adds confidence interval notches around the median.

1. Enable Notches

python ax.boxplot(data, notch=True)

2. Interpretation

Non-overlapping notches between two boxes suggest the medians differ significantly at approximately the 95% confidence level.

3. Notch Calculation

```python

Notch extent approximation

notch_extent = 1.57 * IQR / np.sqrt(n) ```

Patch Artist

The patch_artist keyword enables filled boxes for color customization.

1. Enable Patch Artist

python bp = ax.boxplot(data, patch_artist=True)

2. Access Box Patches

python for patch in bp['boxes']: patch.set_facecolor('lightblue')

3. Required for Fill Colors

Without patch_artist=True, boxes are drawn as lines only and cannot be filled.

Show Statistics

Keywords control which statistical markers appear.

1. Show Mean

python ax.boxplot(data, showmeans=True)

2. Show Fliers (Outliers)

python ax.boxplot(data, showfliers=True) # Default ax.boxplot(data, showfliers=False) # Hide outliers

3. Mean Line Instead of Point

python ax.boxplot(data, showmeans=True, meanline=True)

Whisker Range

The whis keyword controls whisker extent.

1. Default (1.5 IQR)

python ax.boxplot(data, whis=1.5)

2. Custom Multiplier

python ax.boxplot(data, whis=2.0) # Fewer outliers shown

3. Percentile Range

python ax.boxplot(data, whis=[5, 95]) # Whiskers at 5th and 95th percentiles

Exercises

Exercise 1. Create a box plot of three datasets with the following customizations: notch=True, patch_artist=True, showmeans=True with a diamond marker. Use different face colors for each box and set medianprops to display a red line.

Solution to Exercise 1 
import matplotlib.pyplot as plt
import numpy as np

np.random.seed(42)
data = [np.random.randn(100) + i for i in range(3)]

fig, ax = plt.subplots(figsize=(8, 5))
bp = ax.boxplot(data, notch=True, patch_artist=True, showmeans=True,
                 meanprops=dict(marker='D', markerfacecolor='gold', markersize=8),
                 medianprops=dict(color='red', linewidth=2))

colors = ['lightblue', 'lightcoral', 'lightgreen']
for patch, color in zip(bp['boxes'], colors):
    patch.set_facecolor(color)

ax.set_xticklabels(['Group 1', 'Group 2', 'Group 3'])
ax.set_title('Notched Box Plot with Means')
plt.show()

Exercise 2. Create a box plot using showfliers=False to hide outliers, and instead overlay individual data points using ax.scatter with jitter (small random horizontal offsets). Use 200 samples from a standard normal distribution across 4 groups.

Solution to Exercise 2 
import matplotlib.pyplot as plt
import numpy as np

np.random.seed(42)
data = [np.random.randn(200) for _ in range(4)]

fig, ax = plt.subplots(figsize=(8, 5))
ax.boxplot(data, showfliers=False, patch_artist=True,
            boxprops=dict(facecolor='lightyellow'))

for i, d in enumerate(data, 1):
    jitter = np.random.uniform(-0.15, 0.15, len(d))
    ax.scatter(np.full_like(d, i) + jitter, d, alpha=0.3, s=10, color='steelblue')

ax.set_xticklabels(['A', 'B', 'C', 'D'])
ax.set_title('Box Plot with Jittered Data Points')
plt.show()

Exercise 3. Create a box plot with widths=[0.3, 0.5, 0.7, 0.9] for four datasets to show how box width affects visual perception. Use patch_artist=True with graduated blue colors (light to dark). Set whiskerprops, capprops, and flierprops to customize every visible component.

Solution to Exercise 3 
import matplotlib.pyplot as plt
import numpy as np

np.random.seed(42)
data = [np.random.randn(100) * (i + 1) for i in range(4)]

fig, ax = plt.subplots(figsize=(8, 5))
bp = ax.boxplot(data, widths=[0.3, 0.5, 0.7, 0.9], patch_artist=True,
                 whiskerprops=dict(color='navy', linewidth=1.5),
                 capprops=dict(color='navy', linewidth=2),
                 flierprops=dict(marker='o', markerfacecolor='red', markersize=4))

blues = ['#c6dbef', '#6baed6', '#2171b5', '#08306b']
for patch, color in zip(bp['boxes'], blues):
    patch.set_facecolor(color)

ax.set_xticklabels(['Narrow', 'Medium', 'Wide', 'Widest'])
ax.set_title('Variable Width Box Plots')
plt.show()