Bissell (1990) proposed an estimator for the process capability index Cpk
assuming the knowledge whether the process mean p < m or p 2 m is available,
where m is the mid-point between the upper and lower specification limits. Kotz,
Pearn and Johnson (1993) showed that for the index Cpk, the natural estimator
gives values for the standard deviation smaller than those of Bissell's. In this
paper we show that by adding a well-known correction factor to Bissell's
estimator, we obtain an unbiased estimator of Cpk whose standard deviation is
smaller than those given in Kotz, Pearn and Johnson (1993). In addition, we
propose a Bayesian-like estimator which relaxes Bissell's assumption on the
process mean. The distribution of the new estimator is shown to be identical to that
of Bissell's estimator.
關聯:
Communications in Statistics-Simulation, 25(2), 321-329
