In this paper we briefly review Bayesian and frequentist prediction inference for time series, and then advocate the use of guaranteed-content prediction intervals. These intervals are such that their content (or coverage) is guaranteed with a given high probability. They, thus, are more...