Supplier selection based on normal process yield: the Bayesian inference

Abstract

Due to the risk in outsourcing, supplier selection is a critical issue for companies. Considerable evidence shows that among the criteria for selecting a supplier, quality is the most critical factor and the process yield index is an efficient tool for assessing the process quality of the supplier. Although the frequentist approach has been adopted to discriminate the degrees of two yield indices to solve the supplier selection problem, unknown parameters must be estimated from samples, which potentially introduce uncertainty into the statistical testing process. Instead of the frequentist inference, this study proposes using the Bayesian inference to derive the posterior distribution of the ratio of two yield indices. Furthermore, a Markov chain Monte Carlo technique is applied to discern the empirical posterior distribution of the ratio with the aim of discriminating the degrees of two yield indices. The simulations show that the proposed method is not only of reasonable empirical size but tests well in terms of power.