Statistical Control for Atribute Data

Life is not made only by numbers. Many industries are performing testing and inspection focused on aspects like Good – Bad, Go – No Go, Pass – No Pass. These industries are using attributes to define the quality of their products. Several statistical methods were developed for this kind of data. As fundamental idea, these methods are relying on method for numeric data.

Most used statistical methods for attribute data are the control charts, like P chart, NP chart, C, chart, and U chart. P and NP charts are used for defectives, while C and U charts are used for defects. Please note the difference in used terms: defective and defect. A defective product is one which is not acceptable in terms of quality, while for complex products, a defect of a product does not necessarily lead to a defective one.

Charts for defectives	Charts for defects
P chart shows the proportion of defectives in each analyzed subgroup	C chart used when the subgroup size is constant, for the number of defects in each analyzed subgroup
NP chart shows the number of defectives in each analyzed subgroup	U chart used when the subgroup size is not constant, for the number of defects per sampled unit in each analyzed subgroup

Each type of chart is detecting the presence of special causes of variation, indicating either the control or lack of control in process. As is for numerical data, the strongest indication of lack of control is finding an observation which falls outside of +/- 3 sigmas from the mean of the process. Please note that we didn't mention the specification limits; the main use of control charts are to determine variation of the process related to the mean.