Statistical performance of a multicomparison method for generalized species diversity indices under realistic empirical scenarios

•We evaluated the statistical performance of a test for generalized diversity indices.•Our simulations showed that the Pallmann-Scherer test is biased and has low power.•The test had a better performance for Hill numbers than for Tsallis family.•Further improvements in the test should focus on incre...

Full description

Saved in:
Bibliographic Details
Published in:Ecological indicators 2017-01, Vol.72, p.545-552
Main Authors: Butturi-Gomes, Davi, Petrere, Miguel, Giacomini, Henrique C., Zocchi, Silvio Sandoval
Format: Article
Language:English
Subjects:
Bootstrap
Diversity test
Gini-Simpson index
Hill numbers
Shannon-Wiener index
Tsallis entropy
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:•We evaluated the statistical performance of a test for generalized diversity indices.•Our simulations showed that the Pallmann-Scherer test is biased and has low power.•The test had a better performance for Hill numbers than for Tsallis family.•Further improvements in the test should focus on increasing its power. The Pallmann-Scherer test is a promising multicomparison procedure to test statistical hypotheses regarding generalized diversity/entropy indices, such as Tsallis family and Hill numbers (Sq and Hq, respectively), which represent alternative ways of profiling species diversity along a gradient of emphasis on species richness versus evenness in abundance distributions. Given the pressing importance of reliably comparing diversity across ecological communities, and since only a few of such procedures are currently available, knowing its statistical performance under realistic ecological scenarios is of strategic importance. In this paper, we evaluated the performance of the Pallmann-Scherer test using computer simulations of communities following different species-abundance distributions, spatially aggregated as widely observed empirically, and sampled by a commonly used quadrat procedure. We found that the test is very conservative for both Sq and Hq, leading to biased significance levels, with low probabilities of type-I error but high probabilities of type-II error (i.e., low statistical power). Although it should be acknowledged that the current method represents an important starting point, further improvements must be made in order to enhance its power and meet the required standards in comparative studies of diversity.
ISSN:1470-160X
1872-7034
DOI:10.1016/j.ecolind.2016.08.054