|
Statistical Analyses in SingleSubject Experiment Research: Effect Sizes based on NonOverlap Methods
Psychology: Techniques and Application. 2018, 6 (2):
89-99.
DOI: 10.16842/j.cnki.issn2095-5588.2018.02.004
Singlesubject experiment usually yields ecologically valid results. However, the lack of statistical analyses hindered the application of singlesubject experiment and the promotion of the results. Effect sizes based on nonoverlap methods, free from the constraints of assumptions made in parametric statistics, are suitable to apply to statistical analyses in singlesubject experiment research. In recent years, more than 10 methods have emerged, among which 5 methods (namely, extended celeration line (ECL), percentage of nonoverlapping data (PND), improvement rate difference (IRD), nonoverlap of all pairs (NAP), and Tau for nonoverlap with baseline trend control (Tau-U)) are rather typical. In selection of the indicators of effect sizes, decisions should be made based on data characteristics, data trend and the discrimination power, accuracy and statistical power of indicators. For the sake of the comparison, promotion of results and the followup metaanalysis, the future use of effect sizes based on nonoverlap methods should be encouraged.
Related Articles |
Metrics
|