one typically must use a conservative interval.One way to get around this problem is to interpolate between two nested intervals as in Hettmansperger and Sheather (1986) and Beran and Hall (1993), but this method yields answers that are approximate rather than exact. A second approach, which was used by Di Bucchianico et al. (2001) and Frey (2010) in the context of tolerance intervals and by Frey (2013) in the context of prediction intervals for a single future observation, is to select the shortest of several non-nested intervals