CW

Advanced Visualization: Plotly and Interactive Charts

Module 3: VisualizationFree Lesson

Advertisement

Static vs Interactive: The Decision Framework

The choice between static and interactive visualization is a cognitive design decision that affects how viewers process information.

Visualization Decision FrameworkStatic (Matplotlib/Seaborn)✓ Print / PDF / Paper✓ Pixel-perfect control✓ Reproducible figures✓ Small file size (KB)✓ Journal submissions✓ Domain expertsBest for: Academic papers, reportsInteractive (Plotly)✓ Web / Dashboard✓ Hover tooltips✓ Zoom / Pan / Filter✓ Animation support✓ Self-contained HTML✓ Stakeholder explorationBest for: Dashboards, presentationsRendering BackendsMatplotlib: CPU (Agg)Plotly: WebGL (GPU)Plotly: SVG (vector)Decision TreePrint/PDF → Matplotlib| Web/Dashboard → Plotly| Both → Matplotlib static + Plotly HTML≤ 3 dims → Static OK| ≥ 4 dims → Interactive (hover/filter)

Plotly Express: The Grammar of Interactive Graphics

Scatter Plots

import plotly.express as px
import plotly.graph_objects as go
import plotly.subplots as sp
import pandas as pd
import numpy as np

df = px.data.gapminder()

fig = px.scatter(
    df.query("year == 2007"),
    x="gdpPercap", y="lifeExp",
    size="pop", color="continent",
    hover_name="country",
    log_x=True, size_max=60,
    title="GDP per Capita vs Life Expectancy (2007)",
    labels={
        "gdpPercap": "GDP per Capita (USD, log scale)",
        "lifeExp": "Life Expectancy (years)"
    }
)
fig.update_layout(template="plotly_white", font=dict(family="Arial", size=12))
fig.show()

Line Charts

df_canada = px.data.gapminder().query("country == 'Canada'")

fig = px.line(
    df_canada, x="year", y="lifeExp",
    title="Life Expectancy in Canada Over Time",
    markers=True,
    labels={"year": "Year", "lifeExp": "Life Expectancy (years)"},
    template="plotly_white"
)
fig.update_traces(line=dict(color="#2E86AB", width=3), marker=dict(size=8))
fig.show()

Bar Charts

df_tips = px.data.tips()

fig = px.bar(
    df_tips, x="day", y="total_bill", color="sex",
    barmode="group",
    title="Average Total Bill by Day and Gender",
    text_auto=".2f",
    template="plotly_white"
)
fig.show()

Histograms and Distribution Plots

fig = px.histogram(
    df_tips, x="total_bill", color="time",
    nbins=30, opacity=0.7,
    histnorm="probability density",
    marginal="rug",
    title="Distribution of Total Bills (Lunch vs Dinner)",
    template="plotly_white"
)
fig.show()

Kernel Density Estimator

f^(x)=1nh∑i=1nK(x−xih)\hat{f}(x) = \frac{1}{nh} \sum_{i=1}^{n} K\left(\frac{x - x_i}{h}\right)

Here,

  • KK=kernel function (typically Gaussian)
  • hh=bandwidth (smoothing parameter)
  • nn=number of data points

Box and Violin Plots

fig = px.box(
    df_tips, x="day", y="total_bill", color="day",
    points="all", notched=True,
    title="Total Bill Distribution by Day",
    template="plotly_white"
)
fig.show()

fig = px.violin(
    df_tips, x="day", y="total_bill", color="day",
    box=True, points="all",
    title="Violin Plot of Total Bills by Day",
    template="plotly_white"
)
fig.show()
Silverman's Rule of Thumb for Bandwidth
h=0.9⋅min⁡(σ^,IQR1.34)⋅n−1/5h = 0.9 \cdot \min\left(\hat{\sigma}, \frac{IQR}{1.34}\right) \cdot n^{-1/5}

Here,

  • σ^\hat{\sigma}=sample standard deviation
  • IQRIQR=interquartile range
  • nn=sample size

Subplots: Multi-Panel Layouts

Plotly Subplot Layouts2×2 GridScatterBarHistogramBoxspecs=[[None, None]]Secondary Y$cntsecondary_y=TrueInset Plotdomain=[0.6, 0.9]Key Parametershorizontal_spacing: Gap between columns (default 0.1)vertical_spacing: Gap between rows (default 0.1)specs: Grid layout with [[{"type": "scatter"}, {"type": "bar"}]]secondary_y: Enable dual y-axessubplot_titles: Auto-titles for each panelshared_xaxes: Link zoom across rows 
from plotly.subplots import make_subplots

fig = make_subplots(
    rows=2, cols=2,
    subplot_titles=("Scatter", "Bar", "Histogram", "Box"),
    horizontal_spacing=0.12,
    vertical_spacing=0.15
)

fig.add_trace(
    go.Scatter(x=df_tips["total_bill"], y=df_tips["tip"],
               mode="markers", marker=dict(size=6, opacity=0.6), name="Scatter"),
    row=1, col=1
)

day_means = df_tips.groupby("day")["total_bill"].mean().reset_index()
fig.add_trace(
    go.Bar(x=day_means["day"], y=day_means["total_bill"],
           name="Bar", marker_color="#2E86AB"),
    row=1, col=2
)

fig.add_trace(
    go.Histogram(x=df_tips["total_bill"], nbinsx=20,
                 name="Histogram", marker_color="#A23B72"),
    row=2, col=1
)

for day in df_tips["day"].unique():
    day_data = df_tips[df_tips["day"] == day]
    fig.add_trace(go.Box(y=day_data["total_bill"], name=day), row=2, col=2)

fig.update_layout(height=700, width=900, title_text="Multi-Panel Dashboard",
                  template="plotly_white", showlegend=False)
fig.show()

Secondary Y-Axes

months = pd.date_range("2024-01-01", periods=12, freq="MS")
revenue = np.random.uniform(50000, 120000, 12)
orders = np.random.poisson(200, 12)

fig = make_subplots(specs=[[{"secondary_y": True}]])

fig.add_trace(
    go.Scatter(x=months, y=revenue, name="Revenue ($)",
               line=dict(color="#2E86AB", width=3), mode="lines+markers"),
    secondary_y=False
)
fig.add_trace(
    go.Bar(x=months, y=orders, name="Orders",
           marker_color="rgba(162, 59, 114, 0.5)"),
    secondary_y=True
)

fig.update_yaxes(title_text="Revenue ($)", secondary_y=False)
fig.update_yaxes(title_text="Number of Orders", secondary_y=True)
fig.update_layout(title="Revenue and Orders Over Time", template="plotly_white")
fig.show()

Statistical Visualizations

Correlation Heatmaps

numeric_cols = ["total_bill", "tip", "size"]
corr_matrix = df_tips[numeric_cols].corr()

fig = px.imshow(
    corr_matrix, text_auto=".3f",
    color_continuous_scale="RdBu_r",
    zmin=-1, zmax=1,
    title="Correlation Matrix Heatmap"
)
fig.show()

Pearson Correlation Coefficient

rxy=∑i=1n(xi−xˉ)(yi−yˉ)∑i=1n(xi−xˉ)2⋅∑i=1n(yi−yˉ)2r_{xy} = \frac{\sum_{i=1}^{n}(x_i - \bar{x})(y_i - \bar{y})}{\sqrt{\sum_{i=1}^{n}(x_i - \bar{x})^2 \cdot \sum_{i=1}^{n}(y_i - \bar{y})^2}}

Here,

  • xˉ,yˉ\bar{x}, \bar{y}=sample means
  • nn=number of paired observations
  • rxyr_{xy}=correlation coefficient (−1 to +1)

Geographical Maps

# Choropleth Map
df_2007 = px.data.gapminder().query("year == 2007")

fig = px.choropleth(
    df_2007, locations="iso_alpha", color="lifeExp",
    hover_name="country",
    color_continuous_scale=px.colors.sequential.Viridis,
    title="World Life Expectancy (2007)"
)
fig.update_layout(geo=dict(showframe=False, showcoastlines=True))
fig.show()

# Scatter Mapbox
np.random.seed(42)
df_map = pd.DataFrame({
    "lat": 40.7128 + np.random.normal(0, 0.05, 200),
    "lon": -74.0060 + np.random.normal(0, 0.05, 200),
    "value": np.random.exponential(50, 200)
})

fig = px.scatter_mapbox(
    df_map, lat="lat", lon="lon", color="value", size="value",
    color_continuous_scale="Viridis", size_max=15, zoom=11,
    mapbox_style="carto-positron"
)
fig.show()

Animated Charts

df = px.data.gapminder()

fig = px.scatter(
    df, x="gdpPercap", y="lifeExp", size="pop",
    color="continent", hover_name="country",
    log_x=True, size_max=55,
    range_x=[100, 100000], range_y=[25, 90],
    animation_frame="year", animation_group="country",
    title="Global Development Over Time (1952-2007)"
)
fig.show()
Animation Frame PipelineDataFrame SplitRenderAnimateanimation_frame: column to iterate overanimation_group: entity to track across framesPlay/Pause buttons and timeline slider auto-generated

Dashboard-Ready: Updatemenus and Sliders

# Dropdown filter
fig = go.Figure()
continents = df["continent"].unique()
for continent in continents:
    cdf = df[df["continent"] == continent]
    fig.add_trace(go.Scatter(
        x=cdf["gdpPercap"], y=cdf["lifeExp"],
        mode="markers", marker=dict(size=cdf["pop"] / 2e7, opacity=0.6),
        name=continent,
        visible=True if continent == "Asia" else "legendonly"
    ))

fig.update_layout(
    updatemenus=[dict(
        buttons=[
            dict(label="All Continents", method="update",
                 args=[{"visible": [True] * len(continents)}])
        ] + [
            dict(label=c, method="update",
                 args=[{"visible": [i == idx for i in range(len(continents))]}])
            for idx, c in enumerate(continents)
        ],
        direction="down", showactive=True, x=0.17, y=1.15
    )],
    template="plotly_white", height=550
)
fig.show()

# Range slider
fig = go.Figure()
fig.add_trace(go.Scatter(
    x=df_canada["year"], y=df_canada["lifeExp"],
    mode="lines+markers", line=dict(color="#2E86AB", width=3),
    fill="tozeroy", fillcolor="rgba(46,134,171,0.2)"
))
fig.update_layout(
    xaxis=dict(
        rangeselector=dict(buttons=[
            dict(count=10, label="10Y", step="year", stepmode="backward"),
            dict(count=20, label="20Y", step="year", stepmode="backward"),
            dict(step="all", label="All")
        ]),
        rangeslider=dict(visible=True), type="date"
    ),
    title="Life Expectancy with Range Slider", template="plotly_white"
)
fig.show()

Customizing Themes and Export

# Built-in templates
import plotly.io as pio
pio.templates.default = "plotly_white"

# Custom template
custom_template = go.layout.Template(
    layout=go.Layout(
        title=dict(font=dict(family="Helvetica", size=18, color="#1a1a2e"),
                   x=0.5, xanchor="center"),
        font=dict(family="Helvetica", size=12, color="#1a1a2e"),
        plot_bgcolor="white", paper_bgcolor="white",
        colorway=["#2E86AB", "#A23B72", "#F18F01", "#C73E1D", "#3B1F2B"]
    )
)

# Export
fig.write_image("figure.png", scale=3, width=800, height=500)  # 300 DPI
fig.write_image("figure.svg", width=800, height=500)  # Vector
fig.write_html("interactive.html", include_plotlyjs=True)  # Standalone

File Size Optimization

A standalone HTML with include_plotlyjs=True bundles ~3.5 MB Plotly.js. For multiple charts, use include_plotlyjs='cdn' to load once from CDN, reducing per-chart overhead to ~5 KB.

Library Comparison

CriterionMatplotlibSeabornPlotly
Best forFine-grained controlStatistical plotsInteractive dashboards
InteractivityNoneNoneNative
Statistical modelsManualBuilt-inLimited
3D plottingAxes3DLimitedscatter_3d
AnimationFuncAnimationLimitedanimation_frame
Export qualityExcellentExcellentGood (kaleido)
Browser embeddingNoNoYes

Key Takeaways

Summary: Advanced Visualization

  1. Interactive visualization is a cognitive design choice — use Plotly when stakeholders need to explore data; use matplotlib when precision and reproducibility matter
  2. Plotly Express provides the highest-level API — a single function call generates a fully interactive figure with hover, zoom, and legends
  3. make_subplots is the foundation for multi-panel dashboards — combine specs with secondary_y for complex layouts
  4. Statistical visualizations (marginal distributions, correlation heatmaps, pair plots) are built into Plotly Express
  5. Animations via animation_frame create self-contained temporal visualizations without external video tools
  6. Export to HTML for web embedding; export to PNG/SVG via kaleido for print and publications at 300+ DPI
  7. The matplotlib vs plotly decision depends on your medium (print vs web), audience, and need for interactivity

Practice Exercises

  1. Create an interactive scatter plot using px.data.iris() with marginal distributions and a trendline
  2. Build a 2x3 subplot dashboard using the px.data.gapminder() dataset
  3. Create the animated Gapminder bubble chart with custom colorscale and range slider
  4. Create a masked correlation heatmap (upper triangle only) with annotations
  5. Create a choropleth map colored by GDP per capita with scatter_mapbox overlay

Advertisement

←Matplotlib & Seaborn VisualizationData Storytelling→

Need Expert Data Science Help?

Get personalized tutoring, project support, or professional consulting.

Contact Us →View Services

Advertisement