ISI - MIA

Abstract

Several methods have been proposed to improve the performance of quality control chats when small and moderate mean shifts occur. Most schemes are known as “adaptive” because they vary at least one parameter (sample size, sampling interval, and control limits) as functions of the observed sampling statistics during each inspection. Prabhu et al. (1993) and Costa (1994) proposed VSS (Variable Sample Sizes), the most well-known adaptive sample size method available today. In this method, a small pre-established sample is used when the sampling statistic falls within the central line region, and a large pre-established sample is used otherwise. Infante e Rodrigues Dias (2006) proposed a method that uses normal pdf to obtain non-necessarily pre-established sample sizes. We propose a methodology where sample size, for the next sampling occasion, is a truncated binomial random variable where p is related to the probability of occurrence of the sampling statistic, and n corresponds to the maximum sample size possible. To evaluate the effect of randomness in sample sizes on chart performance, we compare the proposed methodology to others available in the literature.