ABSTRACT
The effects of non-normality of the distribution of the measured characteristic, X, on properties of Process capability indices (PCIs) have not been a major research item until quite recently, although some practitioners have been well aware of possible problems in this respect. In his seminal paper, V.E. Kane devoted only a short paragraph to these problems, in a section dealing with 'drawbacks'. 'A variety of processes result in a non-normal distribution for a characteristic. It is probably reasonable to expect that capability indices are somewhat sensitive to departures from normality. The use of Edgeworth expansions to represent mild deviations from normality has been quite fashionable. R.R. Clements proposed a method of construction based on the assumption that the process distribution can adequately be represented by a Pearson distribution. M. Munechika utilized an approximate relation between corresponding percentiles of standardized normal and Gram-Charlier distribution to obtain an approximate relationship between the process PCI and the corrected PCI values.
