Advanced Features Example

This example demonstrates the advanced features of UpSet plots including interactive filtering, statistical analysis, and animated transitions.

Setup

First, let’s import our libraries and create some sample data with engagement metrics:

import altair_upset as au
import pandas as pd
import numpy as np

# Create sample data with metrics
np.random.seed(42)
n_users = 1000

# Generate platform usage data with engagement metrics
platforms = ['Facebook', 'Instagram', 'Twitter', 'LinkedIn', 'TikTok']
data = pd.DataFrame({
    platform: np.random.choice([0, 1], size=n_users, p=[0.3, 0.7])
    for platform in platforms
})

# Add engagement metrics
data['daily_time_spent'] = np.random.lognormal(3, 1, n_users)  # minutes
data['posts_per_week'] = np.random.poisson(5, n_users)
data['engagement_rate'] = np.random.beta(2, 5, n_users)

Basic UpSet Plot with Metrics

Create a basic UpSet plot showing platform usage with engagement metrics:

au.UpSetAltair(
    data=data,
    sets=platforms,
    title="Social Media Platform Usage Analysis",
    subtitle="Interactive analysis of user engagement patterns",
    width=800,
    height=500
).chart

Engagement Analysis

Let’s analyze the engagement patterns across different platform combinations:

# Calculate average engagement metrics for each platform
for platform in platforms:
    platform_data = data[data[platform] == 1]
    print(f"\n{platform} Metrics:")
    print(f"Users: {len(platform_data)}")
    print(f"Average daily time: {platform_data['daily_time_spent'].mean():.1f} minutes")
    print(f"Average posts per week: {platform_data['posts_per_week'].mean():.1f}")
    print(
        f"Average engagement rate: {platform_data['engagement_rate'].mean()*100:.1f}%"
    )


# Find most engaged platform combinations
def get_engagement_metrics(group):
    return pd.Series(
        {
            "users": len(group),
            "avg_time": group["daily_time_spent"].mean(),
            "avg_posts": group["posts_per_week"].mean(),
            "avg_engagement": group["engagement_rate"].mean() * 100,
        }
    )


# Calculate metrics for all combinations
combinations = data.groupby(platforms).apply(get_engagement_metrics).reset_index()

# Sort by average engagement
top_engaged = combinations.sort_values("avg_engagement", ascending=False).head(3)
print("\nTop 3 Most Engaged Platform Combinations:")
for _, row in top_engaged.iterrows():
    active_platforms = [p for p, v in zip(platforms, row[platforms]) if v == 1]
    print(f"\n{' & '.join(active_platforms)}:")
    print(f"Users: {row['users']}")
    print(f"Avg Time: {row['avg_time']:.1f} minutes")
    print(f"Avg Posts: {row['avg_posts']:.1f} per week")
    print(f"Avg Engagement: {row['avg_engagement']:.1f}%")

Programmatic Highlighting

You can programmatically highlight specific intersections using the highlight parameter. This is useful for drawing attention to specific patterns without requiring user interaction.

Highlighting the Largest Intersection

Highlight the intersection with the most users:

au.UpSetAltair(
    data=data,
    sets=platforms,
    title="Social Media Platform Usage - Largest Intersection Highlighted",
    highlight="greatest",
    width=800,
    height=500
).chart

Highlighting the Smallest Intersection

Highlight the intersection with the fewest users:

au.UpSetAltair(
    data=data,
    sets=platforms,
    title="Social Media Platform Usage - Smallest Intersection Highlighted",
    highlight="least",
    width=800,
    height=500
).chart

Highlighting Specific Intersections by Index

Highlight specific intersections by their index (0-based):

# Highlight the first intersection
au.UpSetAltair(
    data=data,
    sets=platforms,
    title="Social Media Platform Usage - First Intersection Highlighted",
    highlight=0,
    width=800,
    height=500
).chart

Highlighting Multiple Intersections

Highlight multiple intersections at once:

# Highlight the first three intersections
au.UpSetAltair(
    data=data,
    sets=platforms,
    title="Social Media Platform Usage - Multiple Intersections Highlighted",
    highlight=[0, 1, 2],
    width=800,
    height=500
).chart

Vertical Orientation

UpSet plots can be displayed in either horizontal or vertical orientation. The choice depends on your use case:

• Horizontal layouts (default) are best for static figures in papers

• Vertical layouts are better for interactive plots that can be scrolled

In vertical orientation, the cardinality (intersection size) is displayed horizontally on the left, and set sizes are displayed vertically on top.

au.UpSetAltair(
    data=data,
    sets=platforms,
    title="Social Media Platform Usage (Vertical Layout)",
    subtitle="Better for interactive exploration with scrolling",
    orientation="vertical",  # Use vertical orientation
    width=700,
    height=900
).chart

The matrix can be sorted in various ways - by cardinality (size), degree, or sets - and this works equally well in both orientations.

Comparing Horizontal and Vertical Layouts

Here’s a comparison showing the same data in both orientations:

# Horizontal orientation (default)
au.UpSetAltair(
    data=data,
    sets=platforms,
    title="Horizontal Layout",
    subtitle="Cardinality vertical (top), set sizes horizontal (right)",
    orientation="horizontal",
    width=800,
    height=500
).chart

# Vertical orientation
au.UpSetAltair(
    data=data,
    sets=platforms,
    title="Vertical Layout",
    subtitle="Cardinality horizontal (left), set sizes vertical (top)",
    orientation="vertical",
    width=600,
    height=800
).chart