Globalization has intensified competition in many markets. To remain competitive, the companies look for satisfying the needs of customers by meeting market requi rements. In this context, Process Capability Indices (PCI) play a crucial role in assessing the quality of processes. In the case of non - normal data there are two general approaches based on transformations (Box - Cox and Johnson Transformation) and Percenti les (Pearson’s and Burr’s Distribution Systems). However, previous studies on the comparison of these methods show different conclusions, and thus arises the need to clarify the differences between these methods to implement a proper estimation of these in dices. In this paper , a simulation study is made in order to compare the above methods and to propose a n appropriate methodology for estimat ing the PCI in non - normal data. Furthermore, it is conclud ed that the best method used depends on the type of distribution, the asymmetry level of the distribution and the ICP value.

---

Keywords:

Approximation to frequency distributions, data transformations, normality, process capability indices, simulation.
DOI PDF