Learning and understanding the Kruskal-Wallis one-way analysis-of-variance-by-ranks test for differences among three...

When several treatment methods are available for the same problem, many clinicians are faced with the task of deciding which treatment to use. Many clinicians may have conducted informal "mini-experiments" on their own to determine which treatment is best suited for the problem. These results are usually not documented or reported in a formal manner because many clinicians feel that they are "statistically challenged." Another reason may be because clinicians do not feel they have controlled enough test conditions to warrant analysis. In this update, a statistic is described that does not involve complicated statistical assumptions, making it a simple and easy-to-use statistical method. This update examines the use of two statistics and does not deal with other issues that could affect clinical research such as issues affecting credibility. For readers who want a more in-depth examination of this topic, references have been provided. The Kruskal-Wallis one-way analysis-of-variance-by-ranks test (or H test) is used to determine whether three or more independent groups are the same or different on some variable of interest when an ordinal level of data or an interval or ratio level of data is available. A hypothetical example will be presented to explain when and how to use this statistic, how to interpret results using the statistic, the advantages and disadvantages of the statistic, and what to look for in a written report. This hypothetical example will involve the use of ratio data to demonstrate how to choose between using the nonparametric H test and the more powerful parametric F test.