CorrPlot (Correlation Plot with Dendrogram) Examples

This page demonstrates the interactive CorrPlot chart type in JSPlots.

• Hierarchical clustering: Dendrogram shows similarity grouping of variables
• Dual correlation display: Pearson (upper right) and Spearman (lower left) correlations in one matrix
• Julia-computed correlations: Use compute_correlations() convenience function
• Flexible clustering: Use cluster_from_correlation() with different linkage methods
• Automatic ordering: Variables reordered by clustering for clearer patterns

Example 1: Financial Metrics Correlation Analysis

Analyze correlations between various financial metrics for companies.

This example demonstrates:

• 10 financial metrics tracked for 100 companies
• Computing both Pearson and Spearman correlations in Julia
• Hierarchical clustering with Ward linkage
• Compare Pearson vs Spearman correlations to detect non-linear relationships
• Variables automatically reordered by clustering to reveal patterns

Example 2: Sales Performance Metrics

Examine correlations between sales KPIs using average linkage clustering.

Features:

• Multiple sales metrics: revenue, units, conversion rate, customer satisfaction, etc.
• Average linkage clustering (different from Ward linkage in Example 1)
• Identify which metrics move together
• Detect leading indicators through correlation patterns

Example 3: Scientific Measurements - Single Linkage Clustering

This example demonstrates single linkage clustering (nearest neighbor).

Features:

• Physical and chemical measurements from laboratory experiments
• Demonstrates 'single' linkage clustering
• Single linkage tends to create elongated clusters
• Compare this dendrogram shape to Ward (Example 1) and average (Example 2) linkage

Example 4: Healthcare Patient Metrics - Complete Linkage

Analyze correlations between patient health indicators using complete linkage clustering.

This example includes:

• Vital signs, lab results, and outcome metrics
• Complete linkage clustering (farthest neighbor)
• Identify risk factor correlations
• Compare linear (Pearson) vs monotonic (Spearman) relationships

Example 5: Advanced CorrPlot - Economic Indicators Across Regions

Compare economic indicator correlations across different geographic regions:

• Three regional scenarios: North America, Europe, Asia-Pacific
• Variable selection: Choose which economic indicators to analyze
• Compare patterns: See how indicator relationships vary by region
• Useful for international portfolio analysis and macroeconomic research

Insights to explore:

• How does GDP correlate with unemployment differently across regions?
• Are inflation-interest rate relationships similar globally?
• Which indicators cluster together in each region?

Example 6: Advanced CorrPlot - Climate Variables by Season

Analyze how climate variable correlations change across seasons:

• Four seasonal scenarios: Spring, Summer, Fall, Winter
• Interactive variable selection: Focus on specific climate factors
• Seasonal patterns: See how temperature-humidity relationships vary
• Demonstrates that correlation structure can change with context (season)

Key observations:

• Temperature-humidity correlations differ by season
• Precipitation patterns cluster differently in summer vs winter
• Solar radiation shows different relationships across seasons

Summary

The CorrPlot chart type provides:

• Hierarchical clustering: Dendrogram shows which variables are most similar based on correlation patterns
• Dual correlation display:
  • Top-right triangle: Pearson correlation (measures linear relationships)
  • Bottom-left triangle: Spearman correlation (measures monotonic relationships, robust to outliers)
• Julia-computed correlations: Use compute_correlations() for both Pearson and Spearman
• Flexible clustering: Use cluster_from_correlation() with different linkage methods (:ward, :average, :single, :complete)
• Automatic ordering: Variables reordered by clustering results for clearer visualization of correlation blocks

Basic Workflow (Examples 1-4)

# 1. Compute correlations
vars = [:var1, :var2, :var3, :var4]
cors = compute_correlations(df, vars)

# 2. Perform hierarchical clustering
hc = cluster_from_correlation(cors.pearson, linkage=:ward)

# 3. Create correlation plot
corrplot = CorrPlot(:my_corr, cors.pearson, cors.spearman, hc,
                    string.(vars);
                    title="My Correlation Analysis",
                    notes="Description...")

Advanced Workflow with Multiple Scenarios (Examples 5-7)

# 1. Create multiple correlation scenarios
scenario1 = CorrelationScenario("Short-term", pearson1, spearman1, hc1, labels1)
scenario2 = CorrelationScenario("Long-term", pearson2, spearman2, hc2, labels2)
scenario3 = CorrelationScenario("Volatility", pearson3, spearman3, hc3, labels3)

# 2. Create advanced CorrPlot with interactive features
corrplot = CorrPlot(:advanced, [scenario1, scenario2, scenario3];
                    title="Advanced Analysis",
                    default_scenario="Short-term",
                    default_variables=["var1", "var2", "var3"],
                    allow_manual_order=true)

Advanced Features (Examples 5-7)

• Multiple Scenarios: Switch between different correlation analyses using dropdown
• Variable Selection: Multi-select box to choose which variables to display
• Manual Ordering: Toggle "Order by Dendrogram" off to enable drag-drop reordering
• Interactive Exploration: Compare different correlation contexts (time horizons, regions, seasons, etc.)

Linkage Methods

• Ward (:ward): Minimizes within-cluster variance, creates compact clusters (Examples 1, 5-7)
• Average (:average): Uses average distance between clusters, balanced approach (Example 2)
• Single (:single): Nearest neighbor, can create elongated clusters (Example 3)
• Complete (:complete): Farthest neighbor, creates compact, well-separated clusters (Example 4)

Use Cases

• Financial analysis: Stock correlations across time horizons (Example 5), portfolio diversification
• Economic research: Compare regional indicator relationships (Example 6)
• Climate science: Seasonal correlation patterns (Example 7)
• Scientific research: Discover relationships between experimental measurements (Example 3)
• Healthcare: Analyze patient metrics and identify risk factor correlations (Example 4)
• Feature selection: Identify redundant variables for machine learning

Interpretation Tips

• Dendrogram: Variables that merge early (low height) are highly correlated
• Pearson vs Spearman: If they differ significantly, the relationship may be non-linear
• Clustering: Reveals natural groupings of related variables
• Correlation blocks: After reordering, look for dark red/blue blocks indicating groups of correlated variables
• Scenario switching: Compare how correlations change across contexts (time, region, season)
• Variable selection: Focus analysis on specific subsets of interest
• Manual ordering: Create custom groupings by sector, type, or hypothesis