Permutational analysis of variance
Permutational multivariate analysis of variance (PERMANOVA), is a non-parametric multivariate statistical test. PERMANOVA is used to compare groups of objects and test the null hypothesis that the centroids and dispersion of the groups as defined by measure space are equivalent for all groups. A rejection of the null hypothesis means that either the centroid and/or the spread of the objects is different between the groups. Hence the test is based on the prior calculation of the distance between any two objects included to the experiment. PERMANOVA shares some resemblance to ANOVA where they both measure the sum-of-squares within and between group and make use of F test to compare within-group to between-group variance. However, while ANOVA bases the significance of the result on assumption of normality, PERMANOVA draws tests for significance by comparing the actual F test result to that gained from random permutations of the objects between the groups. Moreover, whilst PERMANOVA tests for similarity based on a chosen distance measure, ANOVA tests for similarity of the group averages.
Calculation of the statistic
In the simple case of a single factor with p groups and n objects in each group,the total sum-of-squares is determined as:
where N is the total number of objects, and is the squared distance between objects i and j.
Similarly, the within groups sum-of-squares is determined:
where takes the value of 1 if observation i and observation j are in the same group, otherwise it takes the value of zero. Then, the between groups sum-of-squares () can be calculated as the difference between the overall and the within groups sum-of-squares:
Finally, a pseudo F-statistic is calculated:
where p is the number of groups.
Finally, the PERMANOVA procedure draws significance for the actual F statistic by performing multiple permutations of the data. In each such the items are shuffled between groups. For each such permutation of the data the permutation F statistic is calculated. The p-value is then calculated by:
Where is the F statistic obtained from the original data and is a permutation F statistic.
Implementation and use
- Anderson, Marti J. (2001). "A new method for non-parametric multivariate analysis of variance". Austral Ecology. 26 (1): 32–46. doi:10.1111/j.1442-9993.2001.01070.pp.x.
- Anderson, Marti J. (2005). "Permutational Analysis of Variance" (PDF).
- Wheeler, Bob; Torchiano, Marco (2016). "lmPerm: Permutation Tests for Linear Models". Retrieved 2019-02-08.