Exactness and reliability of nonparametric estimators of species richness compared by simulation and field data
When estimating any value, the associated measures of error also have to be estimated. For each of the species richness estimators used above (see Chapter 5.2.3) methods of estimating variance and, hence, standard error are available (BURNHAM & OVERTON 1978, OTIS ET AL. 1978, CHAO 1984, CHAO ET AL. 1992 , CHAO & LEE 1992). However, some of these estimators of variance did not perform well in simulations (OTIS ET AL. 1978, BURNHAM & OVERTON 1979). Moreover, they are not comparable with one another and, thus, introduce an additional source of factors influencing the suitability of the estimators of species number. In order to eliminate this influence, two methods of estimating standard error were chosen, which can be used with all of the species richness estimators. The bootstrap technique to estimate standard error EFRON 1981) was already used for the jackknife estimator of species richness (NICHOLS ET AL. 1998B). Chao also suggests using it with some of her esti-mators (CHAO ET AL. 1996, CHAO ET AL. 2001). A related technique is Tukey's jackknife method(MILLER 1974), which proved its usefulness in estimating standard errors of population parameters (MANLY 1977). In a Monte Carlo study both estimators already proved to be similarly useful in estimating stan-dard errors of point estimates EFRON 1981). However, there is no comparative study of the perfor-mance of the bootstrap and the jackknife method for estimating standard errors with the corresponding methods of the species richness estimators (see Chapter 5.2.3). The aim of this study is to detect the most accurate estimator of standard error for each of the selected estimators of species.