Factors affecting the distribution of wrasses in a fjord system: analysis by generalised linear models

Abstract

Background: Generalised linear models (GLMs, see McCullagh & Nelder 1989) form a modern, flexible and unified approach to statistical modelling. GLMs are of particular value to the analysis of data from ecological surveys, because they allow for deviation from the assumption of normality in the models. Thus one may for instance fit models with Poisson or negative binomial (negbin) error terms and loganthmic link functions. The negbin distribution in particular, frequently applied in ecology, and often gives a good approximation to typically skewed catch data (high proportion of zero catch values, low means). Many of today’s statistical computer packages (e.g. S-Plus) have inbuilt options for GLMs.