For 2-sided normal theory tolerance limits there are two approaches based on content:
- assure that the interval covers at least p proportion of the population with specified confidence
- assure that the interval covers no more than (1-p)/2 from each tail of the population
(cf. Hahn and Meeker [1991, 2017] or Krishnamoorthy and Mathew [2009]). The second approach (introduced by Owen, 1964) is more stringent, producing wider intervals, but directly related to an exact lower confidence bound for process capability index Cpk (cf. Chou, Owen, & Borrego, 1990). Currently, JMP only offers the 1st type. Please add the 2nd type as an option.