A statistical analysis of nonlinear regression models for different treatments for layers

As the cost of research increases, mathematical models become valuable tools to answer research questions. A major application of mathematical modeling is accurate estimation of production performance, growth, and feed consumption for poultry research and production. There are many ways that a given data set can be analyzed, and different models have been proposed to fit those curves. To explore the models available, data were investigated from a study on the effects of a series of balanced dietary protein levels on egg production and egg quality parameters in lying hens from 18 to 74 wk of age. Forty eight pullets were assigned to each of 3 different protein levels. The results clearly demonstrated that balanced dietary protein level was the limiting factor for body weight (BW), average daily feed intake (ADFI), egg weight, and egg production. To test differences of fitted curves, the sum of squared reduction test is used. Using a unique data set with data from individual hens, 6 commonly used models were fitted to hen performance technical data. The resulting statistical inferences from using individual and pooled data were compared. There are only differences in using individual or grouped data in fitting nonlinear models to laying hen response data. For the most important response variables, hen-day egg production, and feed intake, predicted responses are within 0.12 and 0.65%, respectively, throughout the production cycle.