
184 Nonparametric Statistics for Social and Behavioral Sciences
individuals or sets. Whether measures are taken on individual subjects or
subjects are organized into sets, each subject should be independent of any
other subject and each set of subjects should be independent of any other
set. Random selection is required to assure representativeness of the sample
to the population to which one wishes to generalize. Although populations
from which the samples were taken do not need to be normally distributed,
they should be somewhat symmetrical in that differences between pairwise
comparisons are similar with similar continuous distributions.