Page 364 - Six Sigma Demystified
P. 364

344        Six SigMa  DemystifieD

                        Interpretation



                        Discrimination Ratio
                        The discrimination ratio provides an estimate of the number of distinct catego-
                        ries or classes discernible by a measurement system. A value of 2, for instance,
                        indicates that the system is useful only for attribute analysis because it can ef-
                        fectively classify the sample units only as being in one group or the other (e.g.,
                        good or bad). AIAG recommends a value of 3 or more, and typically a value of
                        8 or more is suitable for statistical analysis.


                        Gauge R&R
                        The gauge R&R statistic is usually reported as either a percent of tolerance or
                        as a percent of the total process variation. When expressed as percent of toler-
                        ance, gauge R&R indicates the relative usefulness of the gauge system for de-
                        termining acceptability of the process output. Tolerance, calculated as the
                        upper specification limit minus the lower specification limit, indicates the
                        amount of variation that is permitted. As measurement system error increases
                        relative to tolerance, the chances increase that

                          •  Output whose true value is beyond specifications could be measured as
                             being within specifications and subsequently considered acceptable.
                          •  Output whose true value is within specifications could be measured as
                             being outside specifications and subsequently rejected as unacceptable.

                          When expressed as a percent of process variation, R&R indicates the relative
                        usefulness of the measurement system in control charting and process capabil-

                        ity analysis. The calculation of statistical control limits and the estimate of
                        process capability require a sound estimate of process variation. Subgroups
                        plotted on a control chart are subject to this measurement system error as well.
                        As measurement error increases, so does the chance that

                          •  Subgroups from a controlled process will be incorrectly determined to be
                             out of control.
                          •  Subgroups from an out-of-control process will be incorrectly determined
                             to be from an in-control process.

                          Typical recommendations for R&R, expressed as either percent of tolerance
                        or percent of process variation, are
   359   360   361   362   363   364   365   366   367   368   369