Basic theoretical arguments advocating jackknife-2 as usually being the most appropriate nonparametric estimator of total species richness.
Résumé
A lot of nonparametric estimators of the number of unrecorded species after partial sampling of an
assemblage of species, have been proposed in the literature. Unfortunately, these different types of
estimators provides substantially divergent predictions. While empirical comparisons have failed to
consistently select in favour of one among all these estimators, a new approach, based on more
theoretical ground, has proven that among three of the most commonly used nonparametric
estimators, Chao, Jackknife-1 and Jackknife-2, the latter was the best choice in most cases while
Chao or Jackknife-1 should preferably be restricted to samplings approaching completeness. Here,
I propose an alternative approach, aiming also at discriminating between the same three estimators
on the basis of another theoretical argument: The necessary compliance with the required “rule of
additivity”, according to which, if an assemblage of species is made of several, distinct groups of
species, the estimation of species richness for the whole assemblage should be exactly the sum of
the estimations of richness for each group of species. Referring to this rule of additivity, the
Jackknife series of estimators (and in particular Jackknife-2 when samples remain far from
completeness), proves, once again, being satisfactory in full generality. This strengthens the
estimators of the Jackknife series as being particularly appropriate to evaluate, in most cases, the
number of unrecorded species of a partially sampled assemblage and the corresponding total
species richness of the assemblage.