Comparison of fixed and flexible nonlinear regression models with respect to the inflection point: An example of modelling growth in Denizli chickens raised in different systems

The aim of this study was to determine the individual live weights of 156 mixed-sex Denizli chickens from hatching to 26 weeks of age and to determine which of the growth curve models commonly used in poultry is the most suitable for Denizli chickens by using the individual live weight values. Conventional deep litter system, enriched deep litter system, and conventional multi-layer cage systems were also used in the study. In the nonlinear regression analyses of the weekly live weights of Denizli chickens housed in different rearing systems, fixed sigmoidal models with respect to inflection point such as Gompertz, Logistic, Von Bertalanffy, and flexible functions with respect to inflection point such as Richards, Levakovich, Janoschek were used. The differences between the growth profiles of female and male Denizli chickens housed in multi-layer cage, floor rearing and enriched floor rearing systems were tested by Profile analysis technique, which is one of the multivariate statistical methods, and differences were found between the groups (P<0.05). Male Denizli chickens were found to be heavier than females in each rearing system, and it was found that Denizli chickens reared in the conventional deep litter system had higher average live weights than those in the other two systems (P<0.05). In this study, where the growth performances of Denizli chickens reared in conventional cages, conventional deep litter systems and enriched deep litter systems were compared, it was determined that the most compatible model in terms of goodness of fit criteria was the Gompertz model with three parameters and a fixed inflection point as a result of growth curve analyses. Since the model in question has a fixed inflection point, it may have negative effects on genetic variation in breeding studies, it is possible to say that the Richards model, which is flexible in terms of inflection point and has the best results after the Gompertz model in terms of goodness of fit criteria, can also be used.