Prontise in NonparanIetrics (research, statistics) Charlotte Brasic Royeen, William L. Seaver This paper proposes the use of non- parametric statistical methods as a way to increase the quality and quantity of research in occupational therapy. Assumptions underlying parametric versus nonparametric methods are compared and the con- ditions under which nonparametric methods may be preferred are speci- fied. Finally, a synopsis of equivalent nonparametric procedures for com- mon parametric methods is pre- sented. This article was written by the au- thor in her private capacity. No offi- cial support or endorsement by the Department of Education is intended or should be inferred. R esearch-related activity has steadily increased in occupa- tional therapy (I). It is seen as a crucial element in our continued development. A variety of methods have been proposed to increase both the quantity and the qualIty of the research developed within the profession. For example, Kiel- hofner (2) expounded on the meth- ods and advantages of qualitative research. Christiansen (3) deline- ated the role that a simple, yet COIll- prehensive theory such as occupa- tional behavior can play in research generated by occu pa tiona I thera- pists, implying that quality research IS dependent on an adequate grounding ill theory. Ottenbacher suggested that to improve the quality of research ill our profession, we need to attend critically to the uses and abuses of research procedures (4). He sam- pled 14 articles in the occupatiollal therapy literature that employed statistical tests of hypotheses and found that 5 (35.7%) used the t test inappropriately by reporting re- sults of multiple t tests. If the com- parisol1wise error rate is main- tained at .0 I or .05 for each t test, the overall error rate (called exper- imentwise error) for all the com- parisons is drastically increased. Thus, as aile uses more multiple t Lests, the probability of falsely re- jectillg the null hypothesis on at least one of the t tests becomes quite large (5). For instance, if 10 multiple t tests were run at a .05 comparison wise level of signifi- cance, the overall error rate would be up to .401. As all alternative to this misuse of the t test, Ottell- bac her proposed the use of other parametric statistical procedures, such as the analysis of variance (A OVA), with appropriate post hoc analyses. We propose yet another way to improve the quality of research in occupational therapy: nonpara- metric methods of analyses. A de- .i'leation of the assumptions un- derlying parametric statistics will illustrate that nonparametric procedures hold great promise for use ill the field. Parametric procedures such as the t test are calculated to reOect the characteristics of a sample (sta- tistics) so that one can then make Charlotte Brasic Royeen, N/S, OTR, is an educational research analyst with the Office of Special Education, US. Department of Education, Washington, DC 20202. At the time this paper was written, she was a re- search and computer specialist with the Department of Education, Vir- ginia Polytechnic Institute and State University, Falls Church, VA. Wil- liam L. Seaver, PhD, is Assistant Professor of Statistics, Department of Statistics, Virginia Polytechnic Insti- ILite and State University, Falls Church, VA 22042. The American Journal of Occupational Therapy 191 Downloaded from http://ajot.aota.org on 07/11/2020 Terms of use: http://AOTA.org/terms