Unidade de Investigacao em Eco-Etologia, Instituto Superior de Psicologica Aplicada, Lisboa, Portugal
We present a simulation technique used as an alternative to conventional goodness-of-fit tests, that allows the identification of the classes whose deviations between observed and expected frequencies are significant. This technique is also useful in cases where expected frequencies are very low and the criteria for using two tests are not fulfilled.
We illustrate the use of this procedure with data from social interactions in captive groups of an African cichlid fish.