Process capability and data contamination

Abstract

Purpose: The paper centres on process capability and its relation to data contamination. Process capability may be distorted due to imprecise data. The paper analyses to what extent capability changes reflect problems in data so that the changes can be attributed to data sampling rather than the true performance of the process. This is important because it is usually much simpler to increase the precision of data sampling than the process itself.

Methodology/Approach: The paper has two major parts. In part one, effect of data contamination on the observed process characteristic is analysed. The effect is analysed using data obtained from simulated random drawings and the chi-squared test. In the other part, reaction of capability to data contamination is observed. The capability is measured by a univariate capability index.

Findings: Regarding the sensitivity of the index to contamination, it is different depending on the capability before the contamination. This leads to conclusions about when the company using the index should focus more on the way the data is measured, and when it should focus more on improving the process in question. The analysis shows that if the company is used to high levels of capability and records its drop, it is worth analysing its measurement system first, as the index is at higher levels more sensitive to data contamination.

Research Limitation/implication: The study concerns a single univariate index, and the contamination is modelled with only several probability distributions. Originality/Value of paper: The findings are not difficult to detect, but are not known in practice where companies do not realize that problems with their process capability may sometimes lie in the data they use and not in the process itself.