Analysis of inter-sample distances from BIOLOG plate data in Euclidean and simplex spaces

BIOLOG plates have a potential use in quantitative or qualitative comparisons between physiological activities of microorganisms from different soils or soil horizons. However, the scaling of the data advocated by Garland and Mills (1991) restricts the range of subsequent data analyses because the scaled points lie in a constrained simplex space. This article describes methods for analyzing inter-sample distances from BIOLOG plate data in the usual Euclidean space, using total (quantitative plus qualitative) or quantitative information, and in simplex space, using only qualitative information. The methods are applied to data for three replicates of each of five soils. The quantitative comparisons showed that for some soils the replicates formed distinct groups, but for others they did not. Major differences must be due to real differences in activity levels, which need further study. The qualitative comparisons showed that on the first two axes of a simplex PCA the replicates formed compact and distinct groups, although two of the groups were close together. The C-sources that best discriminate between the samples can be found from the eigenvectors. Attention is drawn to the problem of negative data values. (C) 1999 Elsevier Science Ltd. All rights reserved.