Evaluation of Modified Non-Normal Process Capability Index and Its Bootstrap Confidence Intervals

Abstract

Process capability index (PCI) is used to quantify the process performance and is becoming an attracted area of research. A variability measure plays an important role in PCI. The interquartile range (IQR) or the median absolute deviation (MAD) is commonly used for a variability measure in estimating PCI when a process follows a non-normal distribution In this paper, the efficacy of the IQR and MAD-based PCIs was evaluated under low, moderate, and high asymmetric behavior of the Weibull distribution using different sample sizes through three different bootstrap confidence intervals. The result reveals that MAD performs better than IQR, because the former produced less bias and mean square error. Also, the percentile bootstrap confidence interval is recommended for use, because it has less average width and high coverage probability.