ABSTRACT

Hypothesis testing is essential to the scientic method (Chapter 1). In this process, an investigator quanties the strength of a hypothesis using evidence obtained from data.

Statistical hypothesis testing procedures can be placed into one of two general groups: those that use a null hypothesis (formally introduced in Section 6.2) and those that do not. Null hypothesis procedures can, in turn, be subdivided into three types: (1) parametric, (2) permutational, and (3) rank-based permutational. Parametric methods assume particular distributional characteristics for the population(s) underlying the data. Methods 2 and 3 are often referred to as nonparametric because they have no a priori distributional assumptions, or as robust because they are generally resistant to outliers. Topics associated with these approaches are considered in Sections 6.2 through 6.6. Statistical hypothesis tests/ comparisons that do not require a null hypothesis generally involve Bayesian and/or likelihood-based methods. I discuss these approaches in Section 6.7.