Simultaneous prediction in misspecified regression models
by
Madhulika Dube
Department of Statistics, M.D.University,Rohtak 124001 INDIA

Prediction is one of the most important aspects of regression analysis as it sheds light on the model adequacy; predictive mean squared error often forming a basis for choosing an appropriate model. In practice, however, the underlying model is quite often misspecified due to one reason or the other. Therefore, in this article an attempt has been made to study the impact of misspecification of the model on predictive performance of a number of estimators. Depending on the situations, predictions may be carried out for actual or mean value of the study variable. Recently, however a number of research articles have considered simultaneous prediction of the study variable using a target function, which in fact is a weighted average of actual and average value of the response variable. It is, therefore, proposed to study the problem of simultaneous prediction of actual and average values of the study variable in linear regression models when the model is misspecified. An attempt has also been made to obtain optimal weights that minimize the predictive risks of the estimators for simultaneous prediction.

Date received: July 15, 2003