DETERMINING THE BEST FITTED GROWTH MODEL IN JAPANESE QUAILS: A BAYESIAN APPROACH


Kaya Başar E., Narinç D.

ATLAS INTERNATIONAL CONFERENCE ON RESEARCH IN APPLIED SCIENCES, Barcelona, İspanya, 9 - 10 Temmuz 2022, ss.1-6

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Basıldığı Şehir: Barcelona
  • Basıldığı Ülke: İspanya
  • Sayfa Sayıları: ss.1-6
  • Akdeniz Üniversitesi Adresli: Evet

Özet

For many years, numerous mathematical functions have been employed to model the growth of poultry. The majority of these are asymptotic-mechanistic models. Depending on the biological growth phenomenon of the animal, these models are functions that assume the dependent variable has an approximated asymptotic value while the independent variable is at infinity. The majority of specific growth models are nonlinear regression equations with a sigmoidal structure. The studies conducted on poultry using growth models may be categorized into three groups: "finding of the most fit model", "comparison of various scientific application results using growth models", and "genetic structure of the growth curve". The aim of this study is to determine the most suitable growth model for Japanese quail. In similar studies conducted to determine the most suitable growth curve model for Japanese quails, the frequentist approach was widely employed. However, there has been a recent surge of interest in the use of the Bayesian methodology to modeling studies. In this study, the Bayesian approach was applied as an alternative to the frequentist method for analyzing growth curve data. Richards, Gompertz, Negative Exponential, Broody, and Logistic functions, which are commonly used to model growth in poultry, were used in this study. Normal distribution was assigned as prior distribution for all growth curve parameters. Jeffrey's non-informative prior was allocated to the prior distribution of the variance of the residuals. Bayesian analyzes were performed using the MCMC procedure in SAS 9.4 software. After evaluating the trace plots for each parameter in all models, a chain length of 120000 was assumed. The initial 10,000 samples were removed as the burn-in period. The thinning interval was set to 20, and 5,500 samples were utilized to determine the descriptive statistics of the marginal posterior distributions. The models were compared using the deviance information criterion (DIC), which is a typically used Bayesian approach goodness-of-fit criterion. Based on this criterion, the Richards model with the smallest DIC value was determined to be the most appropriate model for Japanese quail growth samples. The posterior mean value of the parameter  defined as the asymptotic weight parameter of the Richards model was found to be 270.0. The other model parameters, , , and , were estimated to be 0.81, 0.046, and 19.30, respectively. On the basis of the fit criterion, the Richards function was followed by the Gompertz and Broody growth models, respectively.

For many years, numerous mathematical functions have been employed to model the growth of poultry. The majority of these are asymptotic-mechanistic models. Depending on the biological growth phenomenon of the animal, these models are functions that assume the dependent variable has an approximated asymptotic value while the independent variable is at infinity. The majority of specific growth models are nonlinear regression equations with a sigmoidal structure. The studies conducted on poultry using growth models may be categorized into three groups: "finding of the most fit model", "comparison of various scientific application results using growth models", and "genetic structure of the growth curve". The aim of this study is to determine the most suitable growth model for Japanese quail. In similar studies conducted to determine the most suitable growth curve model for Japanese quails, the frequentist approach was widely employed. However, there has been a recent surge of interest in the use of the Bayesian methodology to modeling studies. In this study, the Bayesian approach was applied as an alternative to the frequentist method for analyzing growth curve data. Richards, Gompertz, Negative Exponential, Broody, and Logistic functions, which are commonly used to model growth in poultry, were used in this study. Normal distribution was assigned as prior distribution for all growth curve parameters. Jeffrey's non-informative prior was allocated to the prior distribution of the variance of the residuals. Bayesian analyzes were performed using the MCMC procedure in SAS 9.4 software. After evaluating the trace plots for each parameter in all models, a chain length of 120000 was assumed. The initial 10,000 samples were removed as the burn-in period. The thinning interval was set to 20, and 5,500 samples were utilized to determine the descriptive statistics of the marginal posterior distributions. The models were compared using the deviance information criterion (DIC), which is a typically used Bayesian approach goodness-of-fit criterion. Based on this criterion, the Richards model with the smallest DIC value was determined to be the most appropriate model for Japanese quail growth samples. The posterior mean value of the parameter  defined as the asymptotic weight parameter of the Richards model was found to be 270.0. The other model parameters, , , and , were estimated to be 0.81, 0.046, and 19.30, respectively. On the basis of the fit criterion, the Richards function was followed by the Gompertz and Broody growth models, respectively.