Impact Analysis for Regression Models with Heteroscedastic Errors

Abstract

This article concentrates on the development of the impact measures
for the perturbation strategy of simultaneous omission of an individual
variable and an observation in a regression model with heteroscedastic
errors. The effect of particular perturbation strategy on residual sum of
squares, fitted and predicted values is evaluated. Moreover, the
interpretation of change in the significance of the model and or the
significance of the measured variables due to specific perturbation
scheme is highlighted in this article. A particular data example
furnishes analysts with the over all images of the proposed measures.