We analyzed data on presence or absence of 156 vascular plant species on 14
islets of the Cabrera National Park archipelago and found a significant co
rrelation between rat presence and floristic composition. By contrast, isla
nd area, island height, and storm exposition seem to have no effect on spec
ies composition. We chose a subset of plants favoured or depressed by rat p
resence for a non-linear principal component analysis and evaluated success
fully the usefulness of such a subset in predicting rat presence using a ra
ndomization method. To optimally allocate resources, rat eradication campai
gns should be focused in islets displaying rat-free vegetation type.