Bayesian implementation of Skew-Normal distributions for experimental error in the description of pepper growth

The purpose of this work is to implement and compare three types of Skew-Normal distributions together with the Normal submodel for the experimental error through a Bayesian nonlinear modeling of pepper phenotype growth. The posterior means of the parameters  were shown to be invariant to the experimental error. The Sahu Skew-Normal error showed evidence of null asymmetry for all growth models. In general, all Skew-Normal distributions presented the highest accuracy ( ) in relation to the Normal error. The problematic points stand out for the computational cost in the millions of MCMC iterations and the limitations of the BUGS language. The Skew-Normal distribution of Azzalini and Fernandéz & Steel modeled the asymmetry of the data and provided the best goodness of fit (DIC) and the best precision ( ) for the Gompertz and Von Bertalanffy model in relation to the Normal error.