Bias and Dispersion of Overlap Indices: Results of Some Monte Carlo Simulations

摘要: We have used computer simulations to determine the sampling distributions of four indices overlap or similarity: coefficient community, Morisita's index, Horn's information theory and Euclidean distance. Estimates were systematically biased downward when sample size was small expected values close 1. The standard deviations samples greatest intermediate between 0 1, sizes small. In studies having 25, 50, 100, one could expect error an estimated index similarity fall 0.05 0.10, provided that truly drawn from homogeneous populations. suggest be estimate confidence limits on where hypothesis testing is required. addition, efforts should made develop for which statistical measures dispersion bias can derived analytically.