Multivariate Analysis of Covariance in Morphometric Studies of the Reproductive Cycle

Abstract

A new approach for analysing morphometric data of reproductive cycles is proposed, involving multivariate analysis of covariance of the directly measured variables (e.g., total weight and gonadal weight), with body length being the covariate. Multivariate, univariate, and between-group tests can be used progressively if significant differences have been found previously. Seasonal variation and other factors of interest can be described with predicted means of the model, adjusted for covariate, rendering the use of indices such as condition factor and gonadosomatic index unnecessary. A special design of multivariate analysis of covariance, with a pooled covariate by factor interaction, can be used to test the fundamental assumption of homogeneous slopes (parallelism hypothesis) in the standard multivariate analysis of covariance. Data for an Iberian brackish water cyprinodontid fish are used to demonstrate the proposed method.