COMPARISONS AMONG THREE HIERARCHICAL CLASSIFICATION TECHNIQUES USING SIMULATED COENOPLANES

摘要: Three classification techniques, MINFO, MDISP and CLUSTER were applied to simulated coenoplane data which varied in noise, beta diversity, sampling design. The addition of noise (30%) caused substantial alterations the uniformly sampled symmetrical coenoplanes with low diversity. Beta diversity had a profound effect on all classifications as plot groupings occured along axis greatest variation. Sampling design affected results according degree aggregation among groups plots. Those discrete clusters more effectively classified by methods than those uniform or continuous. Both MINFO performed well for most conditions; former displaying some slight advantage at high diversities. is generally weaker performer overall there are few situations where it could be recommended over MINFO.