Generalized linear modeling methods for selection component experiments (Q750350)

This paper is concerned with generalized linear modelling techniques for the analysis of selection components, and extends the work of \textit{F. B. Christiansen} and \textit{O. Frydenberg} [in S. Karlin and E. Nevo (eds.), Population genetics and ecology, 277-301 (1976); see also \textit{H. Østergaard} and \textit{F. B. Christiansen}, Theor. Popul. Biol. 19, 378-419 (1981; Zbl 0466.92012)] on nested hypotheses about such components. The authors begin by considering model specification and the generalized linear model, and then discuss an iteratively reweighted least squares procedure for estimating model parameters (selection components). Testing nested hypotheses is then conveniently conducted within the framework of the generalized linear model by the analysis of deviance. Resistant fitting is used to assess the adequacy of the best fitting model from a series of models. The authors conclude with a section on methods for overdispersed data, and provide a discussion of their use of ML methods in fitting selection component models.