Comparison of multimodel analyses: Frequentist vs. Bayesian approaches applied to age and growth studies of bony and cartilaginous fishes

Abstract

The objective was to compare the frequentist and Bayesian approaches in the multi-model analysis of age and growth of bony and cartilaginous fishes, in order to evaluate their differences in estimating key parameters such as the asymptotic length (L∞), the growth coefficient (k), and the length at birth (L₀). Five species were analyzed, three cartilaginous (Prionace glauca, Carcharhinus falciformis, and Alopias pelagicus) and two bony (Selene peruviana and Peprilus medius), using the von Bertalanffy, Gompertz, and logistic growth models. The results showed that Bayesian models tended to estimate higher values of L∞ and lower values of L₀ compared to frequentist models, suggesting that the latter may underestimate the maximum attainable size and overestimate the size at birth. Likewise, the growth coefficient (k) was lower in the Bayesian models, reflecting slower but biologically more plausible growth rates. In terms of model selection, the frequentist approach favored the logistic model according to the AICc criterion, while the Bayesian approach favored the von Bertalanffy model according to the LOOIC criterion. These findings highlight the usefulness of Bayesian models in more accurately representing growth dynamics, especially in contexts with limited or biased data.