Monitoring logistic profiles in phase I using robust cluster‐based method

Abstract

In order to monitor some qualitative characteristics, it is necessary to examine the relationship between a binary response variable and one or more descriptive variables. This relationship is referred to as a logistic profile in the literature of statistical process control. Maximum likelihood estimation (MLE) is a commonly used method to estimate the parameters of logistic regression profiles. If there are outliers in a data set, this estimation method will not perform well in estimating parameters accurately. In this paper, we propose a robust cluster‐based method for monitoring logistic regression profiles in Phase I. We evaluate the performance of the proposed method versus the non‐cluster‐based methods using numerical simulation. The results show that the cluster‐based method improves the performance of control charts for monitoring logistic regression profiles and works well in detecting step shifts and outliers. This method is entirely superior to the non‐cluster‐based methods in estimating the parameters in the presence of outliers; and for medium to large shifts, it shows acceptable performance compared to the robust methods.