Statistical notes to “The Berger–Parker index as an effective tool for monitoring the biodiversity of disturbed soils: a case study on Mediterranean oribatid (Acari: Oribatida) assemblages” [Biodiversity and Conservation DOI 10.1007/s10531-006-9137-3

Abstract

As authors responsible for the statistics of the recent paper by Caruso et al. (2007), weregret that in the ‘‘Data analysis’’ of the ‘‘Material and methods’’ section it was notspecified that Berger Parker index values generally should be transformed to allowstatistical analysis. In fact, this index is a proportion: precisely, it is the fraction of totalsampled individuals that is contributed by the most abundant species. Proportion data arebounded between zero and one. Errors are thus approximately symmetric in the centre (0.5)but become very asymmetric at the extremes, which can result in violating normality andhomogeneity of variance (Grafen and Hails 2002; Quinn and Keough 2002). This featurestrongly limits the use of linear models (e.g. ANOVA) for statistical analysis of proportionsand the classical advice to solve this problem is to normalise data by the arcsin transfor-mation (i.e. Grafen and Hails 2002). However, a more general approach is the generalisedlinear model (below GLM; McCullagh and Nelder 1989; Crawley 2005) with a binomialdistribution:PrðY