The problem of testing the goodness-of-fit of stochastic resource apportionment models

We present a Monte Carlo test to distinguish among a general class of stochastic resource apportionment models using replicated species abundance data. In this test, the stochastic nature of resource apportionment models is used to carry out a large number of simulations and thus calculate a p-value describing how well the model fits the data. We clarify an existing method and present a necessary modification to reduce the computing time required without affecting the reliability of the test. The use of the method is demonstrated with regional data from the North American Breeding Bird Survey (BBS). The example we present shows that the method provides a good level of discrimination between models that match the data and those that don't. An algorithm to implement our test, written for the application SAS, is freely available from the authors upon request. Copyright (C) 2001 John Wiley & Sons, Ltd.