00381 …c79f45a4235fae8108ab837f609112

385

Regret Indices and Capability Quantification

FigurÄ™ 11. Conversions Between Data Types

informative regret function and yet still analyze our results by direct analogy with attributes data methods. The key concept here is to rescale observed numerical values of regret in a slightly different way so that they behave approximately like either Poisson or standardized gamma random variables; see Hoadley (1981, 1986), Link and Ripacki (1985), and Bellcore (1986, 1987). Poissonization provides the path in FigurÄ™ 11 from the cost-of-poor-quality-data box to the attributes-data box.

The Poisson rescaled values are called Equivalent Nonconformances, EN, and are defined by:

EN(X) = ER-R(X)/VR    [17]

This EN rescaling gives regrets the characteristic property of both Poisson and standardized gamma distributions that their mean value and their variance are equal. This common value for the mean and variance of EN is called Equivalent Expectancy, EE: