|
Npstat
|
NPSTAT and NPFACT are a subset of programs
developed to carry out computer intensive analysis (such as randomization
tests but not bootstrapping). They provide nonparametric methods
of null hypothesis testing.
The null distribution consists of an empirical distribution based
on the data at hand, which may be any shape, rather than a theoretical
distribution generated by analytical methods (typically normal or
known relation to normal, such as z, t, and F). The statistical
tests are nonparametric in that they a) do not require interval
or ratio scale of measurement, b) do not assume the data are randomly
sampled from any a specific population, and c) do not require that
the sample data, the population from which the data are taken, or
the null distribution of the test statistic, have any particular
shape. More traditional approaches to statistical testing may be
found if you click
here.
Variations of these tests most familiar to researchers include the
Wilcoxon-Mann- Whitney test on ranks and Fisher's exact probability
test for dichotomous data in a 2 X 2 contingency table. Major advances
in computing in the last 15 years have allowed these nonparametric
procedures to be adapted to a greater variety of research applications
involving scores in one-way and factorial designs, multiple regression,
and multivariate analysis.
|
|
|
|