Statistical analyses are usually based on models. However, a model is almost always only an approx- imate description of more complicated processes and hence it is always wrong. Given this inexactness, it is important to study the variation in inference results under modifications of model formulation. In general, we may think of influence analysis as a set of methods for assessing the sensitivity of inferences to perturbations in the data or the model. In particular, the local influence analysis con- siders changes in the log-likelihood scale as measures to evaluate influential perturbations and the analysis is local because such changes are examined locally, in a small neighborhood of the assumed, unperturbed, model.
Local Influence Analysis
SARTORI, NICOLA
2006
Abstract
Statistical analyses are usually based on models. However, a model is almost always only an approx- imate description of more complicated processes and hence it is always wrong. Given this inexactness, it is important to study the variation in inference results under modifications of model formulation. In general, we may think of influence analysis as a set of methods for assessing the sensitivity of inferences to perturbations in the data or the model. In particular, the local influence analysis con- siders changes in the log-likelihood scale as measures to evaluate influential perturbations and the analysis is local because such changes are examined locally, in a small neighborhood of the assumed, unperturbed, model.
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
