I want to report a statistical analysis. I used Kruskal Wallis test followed by Dunn multiple comparison posthoc since that data are non parametric. I am familiar with the sequence for reporting one way ANOVA i. e. F(x, y) = values and the p value should be stated afterward. May I ask if I should follow suit while reporting Kruskal Wallis test? If yes, I still cannot find the F values in this test. Can someone assist me please?
The Kruskal-Wallis test doesn't use the F distribution. The test has a statistic termed H (I think) that one can compute exact p values for (based on permutation of ranks probably) and a chi-square approximation. Usually one would report H and df or chi-square and df.
An alternative is to use ANOVA on the ranks of the raw data (which is a very good approximation in most cases and allows you to use conventional multiple comparisons procedures.
Also - very minor point - data are never nonparametric. This is a property of the statistical model (specifically whether it assumes a particular distribution and hence the parameters of that distribution). Other options like a generalised linear model, transformations and parametric ANOVA or bootstrapping might be useful options to consider.
As Thom S Baguley says, the test should report a chi-square value, degrees of freedom, and p-value. You can report these.
There are a few different ways to report he Dunn post-hoc test. If you have a compact letter display, that is often the most succinct.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis