Species abundance patterns: the problem of testing stochastic models

1. Tokeshi (1990) proposed a goodness-of-fit test to distinguish among a general class of stochastic species abundance models using field data. This test is a good first step, but it is impaired by some shortcomings: there is no adjustment for the number of species (the rejection rate increases with the number of species in a data set); the variance of the generated abundance distributions are not taken into account (data sets with higher variance than a model are not rejected). 2. We propose an alternative Monte Carlo test that overcomes these problems. This is a versatile test that can be used with any stochastic model generating distributions.