Summary: | 碩士 === 銘傳大學 === 應用統計資訊學系碩士班 === 100 === The process capability index is defined by the ratio between the process variation and the specification limits. It is used to measure the ability of a process to produce the products consistent. Traditionally, the index is obtained separately from single quality characteristic variable. In practice, however, several correlated quality characteristic variables should be considered to satisfy the consumers’ requirement. Multivariate normal distribution is conventionally assumed for multivariate process capability indices in several studies. But, the robustness of the multivariate process capability index for non-Gaussian distribution does not be considered. In this study, we will propose a new multivariate process capability index for bivariate non-Gaussian data. The trimmed data area is measured to estimate the dispersion of variables by omitting the extreme values. The robustness of the proposed index will be investigated by simulation method for several kinds of non-Gaussian distributions.
|