Methods to Compute Prediction Intervals: A Review and New Results

The purpose of this paper is to review both classic and modern methods for constructing prediction intervals. We focus, primarily, on model-based non-Bayesian the a scalar random variable, but we also include Bayesian with objective prior distributions. Our follows two lines: general based (approximate) pivotal quantities predictive connection between these types described distributions in (log-)location-scale family. discuss extending data complicated dependence structures as well some nonparametric (e.g., conformal prediction).