NON-PARAMETRIC METHODS IN PSYCHOLOGICAL RESEARCH

In recent years statistical techniques have been developed which do not rely on parameters of a known distribution in making statistical inferences. These have been called non-parametric or distribution-free methods. The main presentations of this subject matter with which psychologists are familiar are probably those by Edwards ( 8 ) , Moses ( 22 ) , Mosteller and Bush ( 2 3 ) , and Siege1 (27). However, there are numerous other treatises in this area (e.g., 14, 21, 31). These techniques have been developed mainly to avoid the need for making numerous assumptions as is the case with parametric techniques, thus providing certain advantages. However, in light of recent research it appears that at least with one parametric technique some of the assumed advantages of non-parametric analyses cited have been premature inasmuch as the assumptions of this parametric technique are not as restrictive as once supposed. The purpose of this paper is to consider effects of failure to meet assumptions of the analysis of variance technique on subsequent tests of significance and to suggest that in this area non-parametric techniques be given very limited use in psychological research.