Definition:Nonparametric Method

From ProofWiki
Jump to navigation Jump to search

Definition

A nonparametric method is a technique for statistical investigation in which no assumptions are made about any parameters.


Examples

Arbitrary Example $1$

Suppose we have random samples from two populations with unspecified cumulative distribution functions $\map F x$ and $\map G x$.

A test for:

the hypothesis $\map F x = \map G x$ for all $x$

against:

the alternative hypothesis that $\map F x < \map G x$ for all $x$

would be a test that is both nonparametric and distribution-free.


Arbitrary Example $2$

Suppose we have random samples from two populations with unspecified cumulative distribution functions $\map F x$ and $\map G x$.

A test for:

the hypothesis $\map G x = \map F x$ for all $x$

against:

the alternative hypothesis that $\map G x < \map F {x - \theta}$ for all $x$, where $\theta$ is some parameter

would be a test that is distribution-free but not nonparametric, as it involves a parameter $\theta$.


Also known as

A nonparametric method or nonparametric test is also known as a distribution-free method, but the latter has a wider application and it is advised not to conflate the techniques.

Some sources hyphenate: non-parametric method or non-parametric test.


Also see

  • Results about nonparametric methods can be found here.


Sources