Meaning
A mathematical normalization procedure converts highly non-normal data sets into a standard normal distribution using a family of three distinct distribution curves. Quality engineers apply the johnson transformation when process measurements exhibit severe skewness or multi-modal behavior that standard techniques cannot normalize. This method enables the accurate use of traditional statistical tools.
Mathematical Selection
The transformation selects the best fit from three families of functions, which cover bounded, unbounded, and log-normal shapes. This flexibility allows it to handle data sets that other methods fail to convert. The selection process relies on iterative algorithms to find the optimal parameters.
Process Control
Using the converted measurements allows engineers to set reliable control limits on process charts. When raw data is non-normal, standard chart formulas can lead to frequent false alarms or missed process shifts. The transformed data provides a more reliable base for monitoring daily operations.
Statistical Power
When a manufacturing process is being qualified for volume production, applying this transformation helps teams make accurate claims about the process capability. A pilot run that looks out of control might simply have an asymmetrical distribution that the transformation resolves. This prevents unnecessary and expensive physical modifications to machinery that is actually functioning correctly.
The capability of the line can be evaluated against the required yield limits with greater confidence once the mathematical bias is removed from the analysis.
