Utility of binary data for analyzing structure in island communities

Abstract

Archipelagoes with structured island communities were generated using computer simulation. Abundance data matrices were then coded to form binary data matrices with binary divisions of 0 (co-occurrence), n/8, n/4, n/2, 3n/4, and 7n/8; where n is the average total number of individuals per island in the archipelago. Matrices were analyzed for community structure using a variance test, for which there are analogous tests for abundance and binary data. Co-occurrence data did not yield test results similar to those for abundance data, whereas some of the "rare/common" data sets did yield similar test results. Additionally, in many cases, the opposite conclusion was reached when using co-occurrence data rather than abundance data. It is recommended that when binary data are collected for community structure analysis, some measure of species abundance (e.g., above/below average) is considered, not simply species presence or absence.