摘要:In this comment on an article by Rasch and Guiard (2004), it is argued that the
selection of nonparametric over parametric tests cannot rest solely on the
characteristic of tests as robust against non-normality. Other violations exist.
It is shown that the t-test, which is known to be virtually immune to violations
of the normality assumption, is highly sensitive to violations of the
independence assumption. It is concluded that a dismissal of nonparametric tests
that is based on the result that tests are robust against normality violations,
may be premature. More research is encouraged that addresses issues of
violations of assumptions. This research should include both parametric and
nonparametric tests.