Bootstrap-Based Truncation Parameter Estimation for Monitoring Nonconforming Processes

Process parameters on the g chart are generally estimated based on conventional methods such as maximum likelihood, Benneyan, or minimum variance unbiased estimators. However, these estimation methods have weaknesses if there are outliers in the data that cause the results obtained to be biased to obtain accurate control limits. As a solution to this problem, truncation parameter estimation was developed as a robust parameter estimation in the presence of outlier data. However, in its development, this parameter estimation still has weaknesses, when there are no nonconforming items observed in the phase I sample, causing a lack of sensitivity in the control limits. As a solution to this problem, the purpose of this research is to develop a bootstrap-based truncation parameter estimation. This research uses simulation study and empirical data application. It is found that bootstrap-based truncation parameter estimation is more sensitive when the data conditions are non-contaminating outliers or with contaminating outliers. In addition, the bootstrap-based truncation parameter estimation method has good performance when compared to conventional truncation parameter estimation.