© 1970 by Biometrika Trust
Robustness of normal tolerance intervals
University of Melbourne
Tolerance intervals based on the assumption of normality are investigated for robustness with respect to departure from normality in kurtosis. Two types of tolerance intervals and two types of robustness are considered. It is shown that, for the situation considered, the asymptotic robustness, as the number of observations tends to infinity, depends only on the form of the nonnormal distribution and the probabilities associated with the interval. It turns out that for a certain range of probabilities this asymptotic robustness is surprisingly good.