Pandas Density Plots: DataFrame.plot.density() & plot.kde() Analysis

Use DataFrame.plot.density() or its alias DataFrame.plot.kde() to generate smooth Kernel Density Estimation (KDE) curves. Unlike histograms that divide continuous variables into arbitrary discrete bins, density plots compute a non-parametric probability density function (PDF), providing continuous, smooth representations of metric distribution, modality, and skewness.

What is Kernel Density Estimation (KDE)? 🔝

Kernel Density Estimation places a continuous mathematical kernel (typically a Gaussian normal bell curve) over every observation point and sums all kernels together into a single continuous curve. Key properties include:

• No Bin Boundaries: Avoids the arbitrary histogram bin placement artifacts where moving bin edges changes the visual shape.
• Total Probability Integral: The total area under the density curve equals 1.0 (100% probability).
• Peak Interpretation: Peaks in the curve correspond to clusters of high concentration (modes) within your numeric variable.

Create a basic density plot with plot.density() 🔝

Both df.plot.density() and df.plot.kde() are interchangeable aliases. Let us examine student scores in Mathematics:

import pandas as pd
import matplotlib.pyplot as plt

my_dict = {
    "MATH": [30, 33, 32, 32, 31, 34, 32]
}
df = pd.DataFrame(data=my_dict)
print(df)

# Plot smooth density curve
df.plot.density(figsize=(6, 3.5), title="Math Marks Density Estimation")
plt.tight_layout()
plt.show()

Expected console output:

   MATH
0    30
1    33
2    32
3    32
4    31
5    34
6    32

Tuning curve smoothness with bw_method 🔝

The bw_method (bandwidth) parameter controls how wide each Gaussian kernel is. Bandwidth balances curve smoothness vs fidelity to individual data points:

• Default (Scott’s Rule): Automated statistical heuristic based on sample size and standard deviation.
• Small bandwidth (e.g. 0.1–0.2): Highly sensitive, showing narrow peaks around individual observations (risks overfitting noise).
• Large bandwidth (e.g. 0.8–1.5): Oversmoothed curve showing broad global trends (may mask bimodal clusters).

# Fine-tuned bandwidth adjustment
df.plot.kde(bw_method=0.3, figsize=(6, 3.5), title="KDE with bw_method=0.3")
plt.show()

Overlay multiple series with line styles and colors 🔝

When comparing distributions across multiple metrics or categories, customize line colors and dash patterns (style) to distinguish overlapping curves:

multi_df = pd.DataFrame({
    "MATH": [30, 33, 32, 32, 31, 34, 32],
    "ENG": [28, 30, 29, 31, 27, 32, 29]
})

# Compare two distributions with distinct line styles
multi_df.plot.density(
    style=["-", "--"],
    color=["#1f77b4", "#d62728"],
    figsize=(6.5, 3.5),
    title="Math vs English Score Densities"
)
plt.show()

Dual-scale comparison with secondary_y and mark_right 🔝

When two series have radically different magnitudes or variance (causing one curve to dwarf the other on a shared Y-axis), place one series on an independent secondary Y-axis with secondary_y=True:

When secondary_y=True is active, Pandas appends (right) to the column legend by default. Control this behavior with mark_right=True or mark_right=False:

multi_df.plot.density(
    secondary_y=["ENG"],
    mark_right=True,
    figsize=(6.5, 4),
    title="Dual-Scale Density Comparison"
)
plt.show()

Customise gridlines, rotation, and font sizes 🔝

Fine-tune grid visibility, axis label rotations, and typography:

df.plot.density(
    grid=True,
    rot=45,
    fontsize=12,
    figsize=(6, 3.5),
    title="Configured Density Plot"
)
plt.show()

Save density plots to file using savefig() 🔝

Export your KDE curves to vector or raster image files for publication:

ax = df.plot.density(figsize=(6.5, 4))
plt.title("Probability Density Export")
plt.tight_layout()
plt.savefig("density_distribution_report.png", dpi=300)
plt.close()
print("Exported density_distribution_report.png successfully!")

Practical workflow: order delivery time comparison across shipping tiers 🔝

In logistics optimization, analyzing transit time distributions reveals service consistency and late delivery risks. Let us simulate 1,000 package deliveries across Standard and Express shipping:

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

np.random.seed(42)
n_orders = 1000

std_days = np.random.normal(5.2, 1.4, n_orders)
exp_days = np.random.normal(2.1, 0.6, n_orders)

df_shipping = pd.DataFrame({
    "Standard_Shipping_Days": np.clip(std_days, 1.0, 12.0),
    "Express_Shipping_Days": np.clip(exp_days, 0.5, 5.0)
})

print(f"Standard Median Days: {df_shipping['Standard_Shipping_Days'].median():.2f}")
print(f"Express Median Days: {df_shipping['Express_Shipping_Days'].median():.2f}")

# Assert statistical integrity
assert len(df_shipping) == 1000
assert df_shipping["Express_Shipping_Days"].mean() < df_shipping["Standard_Shipping_Days"].mean()

# Plot comparative density curves
ax = df_shipping.plot.density(
    style=["-", "--"],
    color=["#2b5c8f", "#d95f02"],
    bw_method=0.35,
    figsize=(7.5, 4.5),
    title="Delivery Lead Time Probability Densities by Tier"
)
plt.xlabel("Delivery Time (Days)")
plt.ylabel("Probability Density")
plt.grid(True, linestyle="--", alpha=0.6)
plt.tight_layout()
plt.show()

Expected console output:

Standard Median Days: 5.24
Express Median Days: 2.14

Exercise: transaction basket value distribution 🔝

Given 1,000 customer transaction values across two retail store branches, perform these tasks:

1. Create a DataFrame with Branch_A (mean $65, std $15) and Branch_B (mean $90, std $25).
2. Assert that Branch_B has a higher mean transaction value than Branch_A.
3. Plot both distributions on a single KDE chart using bw_method=0.3, grid=True, and title "Retail Basket Value Density by Branch".

Exercise solution 🔝

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

np.random.seed(202)
n = 1000
branch_a = np.clip(np.random.normal(65, 15, n), 10, 150)
branch_b = np.clip(np.random.normal(90, 25, n), 15, 200)

df_retail = pd.DataFrame({
    "Branch_A": branch_a,
    "Branch_B": branch_b
})

# Assert statistical expectations
assert len(df_retail) == 1000
assert df_retail["Branch_B"].mean() > df_retail["Branch_A"].mean()

print(f"Branch A Mean: ${df_retail['Branch_A'].mean():.2f}")
print(f"Branch B Mean: ${df_retail['Branch_B'].mean():.2f}")

df_retail.plot.kde(
    bw_method=0.3,
    style=["-", "--"],
    figsize=(7, 4),
    title="Retail Basket Value Density by Branch",
    grid=True
)
plt.xlabel("Basket Value ($)")
plt.ylabel("Density")
plt.tight_layout()
plt.show()

Expected console output:

Branch A Mean: $64.40
Branch B Mean: $90.20

Practice in Google Colab 🔝

Experiment with bandwidth tuning, dual-axis density curves, and retail distributions in our interactive Google Colab practice notebook:

Open in Google Colab View on GitHub
Run the density estimation examples interactively, experiment with bandwidth tuning, test multi-curve comparisons, and verify the practice exercise directly in Google Colab. All sample datasets are embedded directly in the notebook.

Further Reading & Related Tutorials 🔝

• Pandas Plot Overview — Master visualization hub covering all Pandas charts.
• Pandas Histograms — Discrete bin frequencies and cumulative distributions.
• Pandas Box Plots — Five-number summaries and outlier box-and-whisker plots.
• Pandas DataFrame.mean() — Calculate arithmetic means and distribution central tendencies.
• SQL GROUP BY Queries — Aggregate and summarize continuous data records in MySQL.
• Pandas Section Home — Master index of all Pandas tutorials, pillars, and guides.

Reference: Pandas DataFrame.plot.density official documentation.

Subhendu Mohapatra

Author & Instructor at plus2net

I write and maintain practical tutorials on Python, PHP, SQL, JavaScript, HTML, jQuery, and web development at plus2net. The tutorials focus on clear explanations, working examples, and code that readers can test and adapt while learning.

About the author   Watch on YouTube

Connect: