ABSTRACT
Classical statistical methods for constructing confidence intervals and performing hypothesis tests are based on two critical assumptions: that the observations are independent of one another and that they come from the same population (probability distribution). Parametric methods add the third assumption that the observations come from a particular probability distribution, usually the normal distribution. Of these three assumptions, the most important one (in terms of maintaining the assumed confidence or significance level) is the first one, the assumption of independent observations (e.g., Millard et al., 1985).
