Categorical Data Analysis Using The SAS® System, 2nd Edition

7.1. Introduction

Parametric methods of statistical inference require you to assume that your data come from some underlying distribution whose general form is known, such as the normal, binomial, Poisson, or Weibull distribution. Statistical methods for estimation and hypothesis testing are then based on these assumptions. The focus is on estimating parameters and testing hypotheses about them.

In contrast, nonparametric statistical methods make few assumptions about the underlying distribution from which the data are sampled. One of their main advantages is that inference is not focused on specific population parameters, and it is thus possible to test hypotheses that are more general than statements about parameters. For example, they allow ...