ABSTRACT
INTRODUCTION Most of the methods considered so far have been based on the Normal distribution. These may be inapplicable for any of the following reasons:
➤ Data are continuous but not Normally distributed. ➤ Data include censored observations, known only to be above or
below a particular value. ➤ Data consist of two independent groups with population
variances that are unequal. ➤ Data are discrete quantitative and so can only take whole number
values. ➤ Data are qualitative, either ordered or unordered.