non-parametric statistics

views updated

non-parametric statistics A branch of statistical inference which makes no assumptions about the underlying distributional form of variables. While parametric statistics are based upon an ideal hypothetical mathematical form for the data (usually the normal distribution), some statisticians claim that for social science data the assumptions required to do this are hardly ever justified, and consequently that it is always preferable to use non-parametric statistics (see, for example, W. J. Conover , Practical Nonparametric Statistics, 1980
). A well-known example, equivalent to the parametric t-test, is the Mann-Whitney U-test.