Cramér-von Mises Test

A balanced approach test that gives equal weight to all parts of the distribution.

Test Overview

What it measures:

The Cramér-von Mises test is an advanced goodness-of-fit test that measures the average squared difference between empirical and theoretical cumulative distribution functions, weighted equally across the entire distribution. Unlike the Kolmogorov-Smirnov test which focuses on the maximum difference, this test provides a balanced assessment by integrating squared differences across all values, making it more sensitive to distributional differences in the middle ranges.

Key Concept:

The test uses a quadratic measure of discrepancy between distributions, giving equal weight to deviations throughout the entire range. This approach makes it particularly effective at detecting differences in the central portion of distributions while maintaining good overall power.

Advantages:

• • Balanced sensitivity across entire distribution
• • Good compromise between power and robustness
• • Less sensitive to extreme values than maximum-based tests
• • Effective for detecting central distribution differences
• • Stable performance across various departure types

Limitations:

• • Lower power than specialized tests for specific departures
• • Less familiar than other normality tests
• • Requires larger samples for optimal performance
• • May miss subtle tail-specific departures
• • Complex calculation without software

Statistical Foundation

Test Statistic (W²):

The Cramér-von Mises statistic measures the integrated squared difference between the empirical and theoretical cumulative distribution functions.

Cramér-von Mises statistic

= sample size

Fn(x)

= empirical cumulative distribution function

F0(x)

= theoretical (normal) cumulative distribution function

Computational Formula:

Practical computing formula

= standardized, sorted observation Φ((Xi-μ)/σ)

= rank of the sorted observation (1 to n)

Critical Values:

• α = 0.10: W² > 0.347

• α = 0.05: W² > 0.461

• α = 0.01: W² > 0.743

• Larger values indicate departure from normality

Distribution Properties:

Null Distribution

Under normality, W² follows a known distribution with tabulated critical values

Test Sensitivity

Most sensitive to differences in the central region of the distribution

Key Assumptions

• • Observations are independent and identically distributed
• • Underlying variable is continuous
• • Adequate sample size (n ≥ 30-50) for reliable critical values

Reading the Cramer-von Mises Fit Plot

Interpretation Guidelines

Decision Framework:

Normal Distribution (W² ≤ Critical):

• • Data shows good overall fit to normality
• • Balanced assessment across entire distribution
• • Parametric methods are appropriate
• • No significant departures detected
• • Central and tail regions align well

Non-Normal (W² > Critical):

• • Significant overall departure from normality
• • Consider distribution-free methods
• • Examine central distribution characteristics
• • May need data transformation
• • Investigate shape and location differences

Sample Size Guidelines:

Small (n < 50)

Limited power, use with caution. Consider other tests for small samples.

Medium (50-200)

Good performance with reliable critical values and balanced sensitivity.

Large (> 200)

Excellent power with stable test characteristics and reliable results.

When to Use Cramér-von Mises:

Recommended Applications:

• • When you need balanced sensitivity across the distribution
• • As a compromise between maximum-based and integrated tests
• • For detecting central distribution departures
• • When robustness to extreme values is important
• • As part of comprehensive normality assessment

sixsigmacenter.com Implementation Guide

Using sixsigmacenter.com Platform:

Module 3: Understand Your Data Shape (Distribution & Normality) at sixsigmacenter.com completely automates the Cramér-von Mises test calculations, handling the complex integration and providing accurate normality assessment with balanced sensitivity.

🚀 Automated Analysis Process

• • Instant Integration: Platform automatically performs the complex quadratic integration
• • Error-Free Processing: Eliminates manual calculation of squared differences and summations
• • Comprehensive Output: Provides W² statistic, p-value, critical values, and distributional assessment
• • Balanced Testing: Automatically weights all distribution regions equally for comprehensive evaluation

📊 Publication-Ready Results

• • Statistical Summary: Complete Cramér-von Mises test results with detailed interpretation
• • Balanced Assessment: Equal-weight evaluation across entire distribution range
• • Decision Support: Clear guidance based on W² magnitude and statistical significance
• • Publication-Ready: Reports emphasizing balanced normality evaluation

⚡ Platform Advantages

• • Speed: Instant results vs. complex manual integration calculations
• • Accuracy: Eliminates errors in quadratic integration and critical value lookup
• • Balanced Power: Optimized sensitivity across all distribution regions
• • Integration: Seamlessly compares with other normality tests for complete assessment

💡 Expert Tip: The platform automatically determines optimal sample size requirements for reliable Cramér-von Mises testing and provides guidance on interpreting results in the context of balanced distributional assessment.

Assumptions & Limitations

Key Assumptions:

Independent Observations

Data points must be independent with no serial correlation

Continuous Distribution

Designed for continuous variables - discreteness affects performance

Adequate Sample Size

Requires moderate to large samples (n ≥ 30-50) for reliable results

Test Characteristics:

✅ Strengths:

• • Balanced sensitivity across distribution
• • Good overall power characteristics
• • Less sensitive to extreme observations
• • Stable critical values
• • Effective for central departures

❌ Limitations:

• • Complex manual calculation
• • Lower power for tail-specific departures
• • Less familiar than other tests
• • Requires adequate sample size
• • May miss subtle violations

Practical Considerations:

• Balanced approach: Provides good compromise between different types of sensitivity

• Computational complexity: Requires specialized software for accurate calculation

• Sample size dependency: Performance improves significantly with larger samples

• Complementary use: Best used alongside other normality tests for comprehensive assessment

Example Data Layout for Cramér-von Mises Test

Single column of continuous data for balanced normality testing with equal weight across distribution. Sample Size Recommendation: Minimum 30-50 observations for reliable Cramér-von Mises test results - adequate sample size ensures stable estimates of cumulative distribution functions and meaningful goodness-of-fit assessment.

This is how your spreadsheet or CSV file should be structured before you upload it. Match the column names, keep one row per observation, and leave the analysis to the platform.

Single column of continuous data for Cramér-von Mises normality testing. Provides balanced assessment across entire distribution range. Sample Size Recommendation: Minimum 30 observations for reliable testing, preferably 50+ for optimal performance with stable critical values and good power characteristics.