Comparing the performance of various estimators using large model communities

Pages 271-291 : illustrations ; 27 cm

Bibliographical references (pages 289-291)

Bibliografia na stronach 289-291

A computer program was constructed that simulates large species assemblages (28 to 997 species) with various species-rank order distributions and degrees of aggregation of the species. From these model assemblages random samples were taken to study the performance of 14 estimators of species diversity. For 6 of the estimators correction factors are developed. In sufficiently large samples (more than 2/3 of the true species number (TS) sampled) a corrected second order jackknife estimator gave the best results. 18% of the estimates ranged outside TS ± 10%. If fewer species are represented in the sample (but more than 1/3 TS) two newly developed data analytical estimators performed better. Between 23 and 24%, respectively, of their estimates ranged outside TS ± 20%. Crucial to the performance of all of the estimators is the sample size. The minimum sample size for an estimator to work has to contain at least 1/3 of the total species number.

Strony 271-291 : ilustracje ; 27 cm