Nonparametric statistics consists of methods that do not require a population distribution to have a fixed parametric form described by a finite list of unknown parameters. It includes methods for estimation, hypothesis testing, and resampling. Some nonparametric methods are distribution-free under a specified null hypothesis.
Nonparametric does not mean assumption-free. A method may still require independence, continuity, symmetry, invariance under specified transformations, or other conditions for its conclusions to be valid.
