[pp.49-54]

Bias and Variability of Similarity Estimation: Results of Some Bootstrap Simulations

Kazuhiro KATOH （the University of Tokyo）

Abstract:
It has been reported that similarity estimation is biased by such factors as the species richness and species evenness. Bray-Curtis index that is a popular similarity index in ecological community analysis has been criticized for showing large bias. This study was conducted to discuss the validity of the use of Bray-Curtis indices though simulations and comparison with relative Euclidean distance, Morisita's Că, which is reported to show the least bias among major indices, and Pianka's alpha, which is similar to Că in mathematical formula. The result showed that the bias of Bray-Curtis index was much larger than that of Morisita's Că, though the largest bias was observed under high qualitative similarity conditions. The result also showed that variability of similarity estimation with Bray-Curtis index or relative Euclidean distance was smaller than that with the other two indices. Qualitative similarity was an important factor to causing the bias of similarity estimation when Bray-Curtis index or relative Euclidean distance was used. In most previous simulation qualitative similarity was not considered and complete qualitative similarity was assumed. It is the reason that Bray-Curtis index was too severely criticized. From practical viewpoint the use of Bray-Curtis index is a possible option unless high qualitative similarity is expected.

Key Word:
Bootstrap simulation, Bray-Curtis index, qualitative similarity, similarity index, species evenness

PDFダウンロード（会員限定）