A Comparison of Parametric, Nonparametric, and Observation Oriented Modeling Techniques

Many individuals have shown concern with the current statistical procedure employed by most social scientists, namely that of null hypothesis significance testing (NHST). Alternative techniques do exist, such as nonparametric testing, which focuses less on assumptions and uses simpler analyses, and observation oriented modeling (OOM), a new modeling paradigm created by Dr. James Grice as a way to focus on data at the individual level. Using these three techniques (a parametric ANOVA, a nonparametric Quade test, and OOM\u27s ordinal pattern analysis) a dataset collected on the QWERTY effect was analyzed to look into the effect of typability of keypress combinations on pleasantness ratings for words. The analyses revealed that the nonparametric test was more powerful and capable of distinguishing more statistical differences and larger effect sizes than the parametric test. These tests are also compared to the OOM test which provided information about how individuals matched on this expected pleasantness pattern. These incongruent findings indicate a need for a greater understanding of all available statistical techniques and more investigation into the proper use of these techniques