Histogram & Normality Test
Verify data normality using Anderson-Darling test — a prerequisite for Cpk, as non-normal data renders Cpk meaningless.
Paste comma- or newline-separated numbers
Before calculating Cpk, quality engineers must verify that measurement data follows a normal distribution (a core Cpk assumption). The histogram gives an intuitive view of distribution shape, while the Anderson-Darling normality test provides a statistical conclusion — together they form the standard pre-analysis step for process studies. This tool is also used to spot abnormal distribution patterns such as bimodal shapes (possibly mixed data from two machines), skewness (processes with physical bounds), or outliers (equipment faults or measurement errors).
Input 200 part-thickness measurements (range 0.90~1.10 mm) Histogram: unimodal, slightly right-skewed, 1 suspicious outlier (1.15 mm) Anderson-Darling normality test: A² = 1.42 (statistic) p-value = 0.012 < 0.05 → Reject normality; data is not normally distributed → Suggestion: remove the outlier and retest, or apply a Box-Cox transformation → If still non-normal, use non-normal Cpk methods
FAQ
Related Tools
Process Capability Index
Once data passes the normality check, the next step is computing Cpk / Cp / PPM to quantify whether process capability meets targets.
Open toolBox Plot
Histograms show distribution shape; box plots quickly reveal the median, IQR, and outliers — ideal for comparing multiple groups.
Open toolMeasurement System Analysis
If data fails normality, the measurement system itself may be the problem. Gauge R&R separates measurement error from true process variation.
Open tool