Font Size: 

In product quality assurance, many concepts are used based on the assumption of normality of manufacturing processes, but often the analysis of this data is carried out either by skipping some mandatory steps, or by misunderstanding some concepts, or by confusing some.

Thus, the paper argues the need to verify the normality of the data - which can be done either visually or with special tests for the normal law. Only after this stage can the tools specific to the normal law be applied to eliminate outliers. Very often here the misinterpretation is encountered that values ​​located at distances greater than 3 standard deviations (sigma) from the mean would be outliers (i.e. 0.27%). But this is not true: using the Boxplot tool, specially designed to eliminate outliers in the case of normal distribution, the percentage of outliers is 0.7% (i.e. values ​​located at distances greater than 2.7 sigma from the mean).

The mathematical meaning of the concept of the “3 sigma rule” is presented below: checking the stability of (normal) processes using the control sheet. It is emphasized that in many cases it is considered that the control sheet verifies a “3 sigma capability” of the process, which is not true at all. The control sheet verifies the stability of the process, and only if the process is stable can we proceed to the final stage: determining the process capability (potential capability C_p and effective capability C_pk).

The following is the difference between the “6 sigma method”, which is a quality management method, and the “6 sigma capability” which is a product capability performance.

In this way, a review of the errors that can be encountered in statistical analyses of product quality was carried out and explanations were provided for all of them.