Bootstrap

A bootstrap estimate of species richness can be calculated as follows (Smith & van Belle, 1984):

1. Randomly select with replacement n samples from the total available, and calculate:

where pi is the proportion of the n that has species i present.

2. Repeat step 1 a large number of times - say 50 to 200, and calculate the mean estimate of Smax and the variance as follows:

where qij is the proportion of the n bootstraps which hold both species i and j.

The output presents a series of estimates using 1, 2, 3... and so on samples. Thus to get sensible estimates you must ensure that all the data come from samples obtained from the same community.

If Number of Random Selections (R) is set > 1 then the sequence samples are selected R times at random from the complete set of samples and the mean estimate calculated. As this removes sample order effects it is generally to be recommended.

By looking at the progressive change in the estimates it is possible to assess if sufficient samples have been taken to stabilise the estimate.

A simple plot of the way the estimate changes with the number of samples used is shown by clicking on the Graph tab in the results window.