Abstract

[This abstract is based on the author's abstract.] The earliest and most common forms of process capability indices (PCI) assume that the process being examined is normally distributed. Violation of this assumption frequently results in inappropriate or misleading results. While the problem has often been recognized, only recently have researchers focused on developing indices for non-normal data. An alternative PCI based on empirical percentiles is proposed. Current research indicates that empirical cumulative distribution functions are appropriate as models for distributions with as few as 100 observations. This suggests that nonparametric forms of PCIs may be utilized to avoid problems encountered by parametric models.