An Alternative Identification of Influential points in Cox Proportional Hazards Model

Influence diagnostics are essential in statistical modeling as the influential points have large effect on any statistical model. Thus, in this article, the identification of influential points in Cox proportional hazards model is considered. There are several diagnostics approaches in Cox proportional hazards model; these approaches are: score residual, scaled score residual, Lmax statistics, and likelihood displacement. We also propose a new diagnostics approach and compare its performance with the existing ones. It is found that the new proposed influential detection performs equally with the existing methods; it works to identify the influential observation.