Ratio Test Statistics and Multiple Comparisons with the Best for Supplier Selection for a Process with Profiles

Abstract

For the quality characteristics described by linear profiles and with very low fraction defectives, the usual supplier selection strategies might not work because a sample of a sensible size most likely contains zero defectives. This study provided decision procedures and rules that are easy to implement based on the ratio test statistics. For the ratio test statistics, prior studies are restricted to two suppliers; in practice, more than two suppliers competing for an order is a standard. The Bonferroni correction method is employed to avoid error inflation due to multiple testing. The proposed method is compared with multiple comparisons with the best method. Tables of critical values and the number of profiles required are provided for practitioners. When comparing more than two suppliers, the proposed method provides useful information to rank suppliers and select the supplier with superior process capability. Furthermore, the statistical properties of the proposed method are investigated. A comprehensive simulation study is done to compare the power and the sample size requirement. The result suggests that multiple comparisons with the best method is superior in terms of power and sample size requirement. A real data set is collected from a computer producer to illustrate the applicability of the methods.