A Novel Statistical Analysis of Cnidocysts in Acontiarian Sea Anemones (Cnidaria, Actiniaria) Using Generalized Linear Models with Gamma Errors

Abstract

Comparative studies on cnidocysts, involving adequate statistical treatment, are very scarce. Classical statistical tests are frequently used assuming normal frequency distributions of capsule lengths, but many distributions are non-normal in acontiarian sea anemones. A traditional choice in these situations are non-parametric tests, although they are not as powerful as parametric tests. An extension of classical methods was developed by some authors; these models, called Generalized Linear Models (GLM), can be used under certain conditions with non-normal data. In view of the properties of our data, that are positive, skewed and with constant coefficient of variation, a Generalized Linear Model with gamma distribution and inverse link function was chosen to analyse the cnidae of acontia from the species Haliplanella lineata, Tricnidactis errans and Anthothoe chilensis. Graphical analysis of residuals showed that these assumptions were reasonable. This method allowed us to avoid transformation of dataset and controversial cases in the limit of significance level. For this task, appropriate subroutines in GLIM language were written. In all cases highly significant differences were found between the specimens considered for every species and nematocyst type (b-rhabdoids, p-rhabdoids B1b and p-rhabdoids B2a).