Statistical evaluation of geobotanical and biogeochemical data by discriminant analysis.
Abstract
Field studies in sclerophyllous woodland in the Eastern Goldfields region of Western Australia showed that it was not visually evident whether any particular plant communities were associated with a particular geological substrate. Discriminant analysis was applied to data from ash analyses of the bark of Eucalyptus lesouefii (the most common plant in the study area) on 63 sampling sites, and data on the distribution of 34 plant species, in an attempt to correlate them with substrates of known composition. Neither of these sets of data required the use of all the variables for maximum discrimination. The plant-distribution data gave 93.2% discrimination and the bark-analysis data 73.0% discrimination. The superiority of the plant-distribution discrimination data was due partly to the greater number of discriminants used.