TITLE: Identifying Bad Leverage Points in Logistic Regression Model Based on Robust Deviance Components
KEYWORDS: Logistic regression, High leverage, Bad leverage, Deviance Residuals
ABSTRACT: High leverage points are outliers in X variables. In logistic regression, high leverage points can either be good or bad leverage points but bad leverage points cause severe inferential effect. Diagnostic methods such as Distance from the Mean (DM) and Robust Logistic Diagnostic (RLGD) are some recent diagnostic tools used to identify high leverage points. These methods take into account the distance in X without determining whether these high leverage points are good or bad. Therefore, in this study we propose a diagnostic method to identify only the bad leverage points. This new detection method is based on deviance component and is referred as Robust Deviance Component (robDEVC) method. Numerical example and simulation study show that the proposed method always ensures only the bad leverage points are correctly identified.
