On the Estimation of Process Parameters in the Taguchi's Approach to the On-line Control Procedure for Attributes

Under the model proposed by Nayebpour and Woodall [5] for Taguchi's on-line control procedure for attributes, estimators for the process parameter vector are derived both from the Classical (maximum likelihood) and Bayesian standpoints. The likelihood function is generated by the detection time of the first defective item under the control procedure. Under the Classical standpoint, a case of nonidentifiability is disclosed. Under the Bayesian standpoint, posterior probability distributions for the process parameters are determined by taking into account independent beta prior distributions.