Fig. 1.8 Many useful multivariate analytical procedures convert (a) a matrix of samples, in this case
the columns are abundances of a suite of taxa (the rows in the table), or (b) other variables into a matrix
of similarities or dissimilarities. This summarises differences across all variables between each pair
of samples into a single distance, or measure of difference between the pair. This information can be
presented graphically in many ways. (c) Non-metric MultiDimensional Scaling (nMDS) plots illustrate
similarity among samples in two- or three-dimensional space, using the ranked distances. Points closer
together on the plot are more similar and vice versa. (d) Clustering attempts to identify ‘natural’ groups
of samples by identifying which are more similar to each other.