Multivariate Normal Test Calculator

Evaluate multivariate normality with multiple tests, matrix diagnostics, transformations, outlier detection, visual checks, detailed interpretations, and exportable results for confident statistical decisions and reporting.

Results summary

Warnings
    Test Statistic Degrees of freedom P-value Decision

    Dataset summary

    Recommended interpretation

    Numerical stability

    Descriptive statistics

    Variable Count Mean Median Std. dev. Variance Minimum Maximum Range Skewness Kurtosis

    Matrix results

    Mean vector

    Covariance matrix

    Correlation matrix

    Inverse covariance matrix

    Eigenvalue summary

    Matrix diagnostics

    Multivariate outlier analysis

    Observation Mahalanobis D² Chi-square cutoff Percentile Status

    Graphs and diagnostics

    Dataset and test settings

    Use the first row for variable names. Each later row represents one observation.

    Formula used

    The calculator estimates a mean vector and covariance matrix. Mahalanobis distance measures each observation’s joint separation. Larger values may indicate multivariate outliers.

    Mean vector: μ = (1/n) Σxᵢ Sample covariance: S = (1/(n−1)) Σ(xᵢ−μ)(xᵢ−μ)ᵀ Mahalanobis distance: Dᵢ² = (xᵢ−μ)ᵀS⁻¹(xᵢ−μ) Mardia skewness: b₁,p = (1/n²) ΣΣ[(xᵢ−μ)ᵀS⁻¹(xⱼ−μ)]³ Mardia kurtosis: b₂,p = (1/n) Σ[Dᵢ²]²

    The null hypothesis states joint normality. Small p-values challenge that assumption. Visual diagnostics should support every final conclusion.

    How to use this calculator

    1. Paste numeric data with observations arranged in rows.
    2. Keep variables in separate columns with clear names.
    3. Select tests, preprocessing, precision, and significance level.
    4. Run the analysis and inspect all warnings carefully.
    5. Review tests, matrices, plots, and detected outliers.
    6. Export the report or cleaned dataset when finished.

    Example data table

    HeightWeightAge
    1706528
    1655931
    1808035
    1757229

    Assumptions and limitations

    Multivariate tests react differently to sample size. Outliers can dominate several statistics. Always combine formal and visual evidence.

    Browser calculations use stable numerical approximations. Results may differ slightly from specialist software. Confirm high-stakes analyses independently before reporting.

    Uploaded files remain inside your browser session. This page sends no dataset server-side. Local processing helps protect confidential information.

    Frequently asked questions

    What is multivariate normality?

    It describes joint normal behavior across variables. Every linear combination should appear normally distributed. This assumption supports many multivariate procedures.

    Which test should I prefer?

    Henze-Zirkler provides broad general-purpose detection. Mardia separates skewness and kurtosis departures. Review several methods before deciding.

    Does a large p-value prove normality?

    No test can prove exact normality. A large p-value shows insufficient contrary evidence. Diagnostics still remain important for interpretation.

    Why does sample size matter?

    Small samples may hide real departures. Large samples detect minor harmless deviations. Practical context should guide final decisions.

    What causes a singular covariance matrix?

    Perfectly related variables create singularity. Constant columns cause the same issue. Too many variables can also destabilize inversion.

    Should outliers be removed automatically?

    Automatic deletion can distort valid findings. Investigate measurement and data-entry causes first. Report every exclusion transparently and carefully.

    When should variables be standardized?

    Standardization helps when scales differ greatly. Mahalanobis distance already uses covariance scaling. Transformations may still improve distribution shape.

    Can missing values be analyzed?

    Complete-case analysis removes incomplete observations. Mean imputation preserves rows but reduces variability. Advanced studies should use principled imputation methods.

    Why do tests disagree?

    Each method targets different distribution departures. Power also changes with dataset dimensions. Use disagreement as a diagnostic signal.

    Related Calculators

    Average Calculator StatisticsGeometric Mean CalculatorInter Quartile Range CalculatorLower Quartile CalculatorMaximum CalculatorMean Calculator StatisticsMedian Calculator StatisticsMidhinge Calculator StatisticsMid Range Calculator StatisticsMinimum Calculator Statistics

    Important Note: All the Calculators listed in this site are for educational purpose only and we do not guarentee the accuracy of results. Please do consult with other sources as well.