Prediction in Random Coefficient Regression Models

Much attention has been given to the problem of predicting future observations for some individual within a random coefficient regression (RCR) model, using the previous observations on that individual as well as the information from the rest of the data material. In this paper, the literature on this subject is critically reviewed and new methods of linear prediction are proposed for the general RCR model. Exact results are derived for the mean squared errors of some predictors in a special case, but this is not possible in the general RCR model when its parameters are not known. In this model, the old and new predictors are compared in a simulation study, and further illustrated by prediction in a medical data material.