Improving Prediction Intervals: Some Elementary Methods

In this article, we consider the problem of constructing prediction intervals for predicting future values of a random variable drawn from a sampled distribution. Two elementary prediction interval calibration methods are proposed to improve the coverage accuracy of prediction intervals. One uses the Box-Cox normal transformation to derive exact prediction intervals, whereas the other suggests an exponential distribution transformation to provide prediction intervals with zero coverage error. Both methods are shown to attain very accurate coverage via numerical comparison Studies.