Statistical methods that do not rely on the assumption of a known population probability distribution function for their validity are called Distribution-free Statistical Methods (also called nonparametric statistical methods). This includes many methods for presenting statistics in graphs, such as histograms, cumulative distribution functions, correlograms, etc. It also includes the calculation of most sample statistics, such as sample means, sample variances, etc. The method is no longer distribution-free when the validity of the method depends on an assumption that specifies the form of the population probability distribution, or specifies that the population distribution comes from a family of population probability distribution functions that are specified except for a finite number of parameters. Then the method is in the class of parametric methods. Usually, the distributional assumption is the assumption that the population has a normal probability distribution, with one or both of the parameters (mean and variance) unspecified.
|Number of pages||9|
|Journal||Wiley Interdisciplinary Reviews: Computational Statistics|
|State||Published - Sep 2009|