Assessing Influence on Partially Varying-coefficient Generalized Linear Model

Abstract

In this paper we discuss estimation and diagnostic procedures in partially varying-coefficient generalized linear models based in the penalized likelihood function. Specifically, we derive a weighted back-fitting algorithm to estimate the model parameters using smoothing spline. Moreover, we developed the local influence method to assess the sensitivity of maximum penalized likelihood estimators when small perturbations are introduced into the model or data. Finally, an example with real data of ozone concentration is given for illustration.

Keywords

• exponential family
• maximum penalized likelihood estimators
• likelihood displacement
• semiparametric models
• weighted back-fitting algorithm