Anpassung eines Fixed Regression Modells für die tägliche Zunahme von Fleckviehbullen mit Hilfe von Informationskriterien

Abstract. Title of the paper: Fitting a fixed regression model for daily gain of bulls using information criterion
In this study the model choice is demonstrated exemplarily on data of 6405 Czech Simmental bulls using information criterion. Per bull up to 8 observations were available for the trait daily gain. Because the animals showed different age on control day, the expected gain curves were described in the population and within the herd*year*season-classes by second, third or fourth order Legendre polynomials of age. For optimization of the fixed effects and to choice the covariance structure of the repeated records the information criteria of Akaike (AIC), the Bayesian criteria (BIC) and the ICOMP-criteria, developed mainly from Bozdogan, were used. Within and over all covariance structures AIC selected generally the most complex model. On the other hand, BIC and ICOMP favoured a model with second order polynomials of age nested within the head*year*seasonclasses. All criterion selected models with nested second order polynomials within the herd*year*season-classes in comparison to models with non-nested polynomials of age.