What non-parametric test should I use?


New Member
Hi all,

I am looking for some help on finding an appropriate test to analyse the variance between groups.

The data I am assessing is:

- whether there is a difference between the downstream distribution of physical habitat between 2003, 2012 and 2016. I have 14 data points for each year recorded at the same places each year.

- whether there is a difference between shear strength and the composition of 5 different types of vegetation at 4 sites (the same two sites at summer and winter)

In both cases data is not normally distributed and the distributions are not a similar shape. So this rules out parametric testing, kruskal-wallis tests and mann-whitney test, right?

Any help would be much appreciated.


To the best of my knowledge, no standard tests exists for the described situation because you have both within-subjects variation (repeated measures) and between-subjects variation (vegetation types and sites). For example, Friedman's Repeated Measures ANOVA is non-parametric but assumes only the repeated measures component...


Various extensions of the Friedman test have been published in research articles but the easiest approach for you is to write down the simplest model capturing everything, with the residuals matching what you are seeing in the data. Then you can test any coefficient in the model using a likelihood ratio test or bootstrap. You can write down your model along the lines of Friedman's model. The simplest form should be tried because you have so little data.