4 2 4 - 0 Three replicate sediment cores were taken for meiofaunal analysis on each occasion, and nematodes identified and counted. This analysis considers onl y the mean nematode abundances across replicates and season (no seasonal differences were evident in a more detailed analysis), so the data matrix consists of 182 species and 19 samples. Instead they appear to relate to variables such as sediment type and organic content, and these links are discussed further in Chapter 11.
Fig. 2 isa map of the sample locations, sites 1-58 (site 30 not sampled). -~ ~ ~ -'3 ~ 73 ~ ~ =:s =3 ~ ~ :;3 =3 =3 Fig. 3 shows the results of a hierarchical clustering using group-average linking on data sampled during April 1974. The raw data were expressed as nu mbers per cubic metre for each of 24 holozoo plankton species, and Bray- Curtis similarities calculated on ,J,J-transformed abundances. From the resu lting dendrogram, Collins and Williams select the four groups determined at a 55 % similarity level and characterise these as trlle eslHarine (sites 1-S, 10, 12), eshwrine and marirre(9, 11, 13-27, 29), eHryhaline marine (28, 31, 33-35, 42-44, 47-50, 53-55) and slenohaline marine (32, 36-41, 45, 46, 51 , 52, 56-58).
For each species abundance, subtract the mean count and divide by the standard deviation over all samples fo r that species. This makes the variance of samples along all species axes the same (= 1) so all species are of potentially equal importance in determining the principal components. This normalised analysis is referred to as cOrTelation-based peA rather than the covariancebased peA obtained when the data is not normalised ' Chopt«4 page~ (the terminology comes from whether the algebraic extraction of eigenvalues and eigenvectors takes place on the correlation or covariance matrix between species).