[R-sig-eco] bootstrapping diversity indices
Kay Cecil Cichini
Kay.Cichini at uibk.ac.at
Tue Jul 13 14:41:43 CEST 2010
hello all,
i want to know whether pooled invaded quadrats (n=6) show decreased
diversity compared to pooled uninvaded quadrats (n=6) - or in other
words, whether species are lacking from invaded quadrats.
to yield confidence limits i bootstrapped pooled absence/presence
species data from 6 quadrats using inverse simpson diversity (the
number of equally-common species required to give a particular value).
i took the boot-samples from a vector where species identifiers were
multiplied with nr. of occurences, and calculated simps.div. from the
tabulated bootstrapped vectors 1000 times.
the boostrap estimates of diversity indices are known to be negatively biased.
i get bias of magnitude -4 to -9 which i use for the adjustment of the
diversity estimates from the observed pooled values and wonder if i
should be concerned about the magnitude of bias or not. histogramms of
boot-samples look ok...
i also used simpson index with small sample corr. (with form:
sum(x)*(sum(x)-1)/sum(x*(x-1))), where the bias is small but the
estimates from the observed values tend to correct for high amounts
(large difference to uncorrected estimates).
any pointers on how to deal with this would be greatly appreciated,
kay
More information about the R-sig-ecology
mailing list