box-cox transformation

**Box-Cox transformation** is the **power-transform method that finds a lambda value to reduce skew and approximate normality for positive-valued data** - it is one of the most common preprocessing steps for capability analysis on right-skewed metrics. **What Is Box-Cox transformation?** - **Definition**: Family of power transformations parameterized by lambda, including log transform as a special case. - **Best Fit Domain**: Most effective for strictly positive data with moderate right skew. - **Parameter Selection**: Lambda chosen by maximizing likelihood or minimizing normality test statistics. - **Output**: Transformed data with improved symmetry and more stable variance behavior. **Why Box-Cox transformation Matters** - **Capability Accuracy**: Improves validity of normal-based indices on skewed process metrics. - **Tail Control**: More accurate upper-tail estimation for defect-risk evaluation. - **Workflow Simplicity**: Widely supported in SPC software and quality toolchains. - **Interpretability**: Power-family behavior is transparent and easier to explain than complex mappings. - **Model Stability**: Often reduces influence of extreme outliers on sigma-based metrics. **How It Is Used in Practice** - **Precheck**: Verify data positivity and remove special-cause outliers before fitting lambda. - **Lambda Fit**: Estimate optimal lambda and validate transformed distribution with probability plots. - **Capability Calculation**: Transform specs and compute indices in transformed domain with back-context reporting. Box-Cox transformation is **a dependable workhorse for handling skewed SPC data** - correct lambda selection often turns unstable capability conclusions into statistically sound ones.