A0307
Title: Censored experiments for computing the average run length
Authors: Sungim Lee - Dankook University (Korea, South) [presenting]
Johan Lim - Seoul National University (Korea, South)
Abstract: The purpose is to introduce a simple and efficient method for computing the average run length, commonly used to measure control chart performance. Generally, a large positive number is assumed, and then many run lengths are taken to compute the average run length in a simulation which is very time-consuming. Moreover, deleting cases with a run length larger than the predetermined maximum necessarily causes bias in computing the average run length. The method suggests this step is unnecessary if the predetermined run length can be represented by type I censored data. Assuming memoryless run lengths, the mean and standard deviation are estimated. Traditional Monte Carlo simulation is compared with the proposed procedure, including Markov chain approximation or integration methods, depending on the control chart type (Shewhart-type, EWMA, or CUSUM). The results show that the proposed methods outperform traditional methods, demonstrating the approach's applicability across various scenarios.
