This paper investigates the local influence assessment in Poisson generalized linear models. The assessment of local influence in generalized linear models is studied by the approach of Cook (1986). So he only deals with the local influence on the regression coefficients which are not resistant to masking and swamping effects.
Suárez and González (2000) propose a new measure to detect locally influential data under perturbations to variance. Based on this measure, in this paper we propose a locally influential measure to mitigate these difficulties. We demonstrate the need of this measure. An example shows the effectiveness of the proposed
method.
keywords: generalized linear model, masking and swamping, local influence.