group, and variance n1n2(n1n21):12. For
group sizes smaller than 20 the normal approximation
is still good or acceptable in many cases.
For samples with relatively many ties, however,
the exact nonparametric probability distribution
is different from data set to data set because it
would not only depend on the group sizes but
also on the number and the distribution of ties
over the two groups. The calculation of such
exact probability distributions would be extremely
cumbersome. In such cases a good solution is
found in the normal approximation extended with