A diagnostic for the general linear model : an application to Time Series

Abstract

An outlier is an observation which is thought to be unusual. The detection of such extreme values is an important issue. Developing a model based on data containing even a single outlier can seriously bias population inferences. The valuable role marginal and conditional residuals play in assuring model robustness is well established in the context of general linear models. The purpose of this thesis is to explore the potential of a statistic, developed for general linear models which incorporates both marginal and conditional residuals, as a diagnostic tool for time series and longitudinal data.