Predictive Influence in the Log Normal Survival Model

We discuss case deletion diagnostics for prediction of future observations in the log normal survival analysis model. The point of view taken is that prediction is the primary inferential goal in a survival analysis setting and that particular observations in the sample may be influential with regard to that goal. We thus consider the Kullback-Leibler divergence as a measure of the discrepancy between predictive densities based on full and case deleted samples. A large Kullback-Leibler number for a particular case is an indication that deletion of that case may result in substantially different predictive inferences.