Diagnóstico de influência local no modelo de calibração ultraestrutural com réplicas

Abstract

This paper aims to present the local influence diagnostic methodology of Cook (1986) along with the selection of the appropriate perturbation schemes proposed by Zhu et al. (2007) in ultrastructural calibration model with replicas. The methodology of Cook (1986) will be used to investigate the robustness and sensitivity of model, where the adopted perturbation schemes were weighting cases and response variables. Perturbing the model and/or data in an arbitrary way can lead to miss interpretations of diagnostic analysis and wrong conclusions. Therefore, this study will evaluate the induced perturbations according to the methodology of Zhu et al. (2007) and if the perturbations are not suitable, we will propose a new way of perturbing the model or the data. As an application it was considered a data set with balanced repeated replication to evaluate which levels and laboratories exercise a disproportional effect on inferences made in the model.