The standard deviation structure as a new approach to growth analysis in the farmed Pacific white shrimp, Penaeus vannamei (Decapoda, Penaeidae)

Publisher Website
Google Scholar
Abstract
In the present study, weight-at-age data of reared Pacific white shrimp, Penaeus vannamei Boone, 1931, were analysed under four different assumptions of variance (observed, constant, depensatory, and compensatory variance) in order to analyse the robustness for selecting the right standard deviation structure to parametrize the Von Bertalanffy, Logistic and Gompertz models. Selection of the best model and variance criteria were obtained based on the Bayesian information criterion (BIC). According to the results of the BIC, the observed variance in the present study proved to constitute the best way to parametrize the three above-listed growth models, and the Von Bertalanffy model appeared to be the best to represent the growth curve found. This is an asymptotic sigmoid curve with an inflection point. Based on these results, using the observed error structure to calculate the growth parameters in multi-model inference analyses is recommended. In the present study, weight-at-age data of reared Pacific white shrimp, Penaeus vannamei Boone, 1931, were analysed under four different assumptions of variance (observed, constant, depensatory, and compensatory variance) in order to analyse the robustness for selecting the right standard deviation structure to parametrize the Von Bertalanffy, Logistic and Gompertz models. Selection of the best model and variance criteria were obtained based on the Bayesian information criterion (BIC). According to the results of the BIC, the observed variance in the present study proved to constitute the best way to parametrize the three above-listed growth models, and the Von Bertalanffy model appeared to be the best to represent the growth curve found. This is an asymptotic sigmoid curve with an inflection point. Based on these results, using the observed error structure to calculate the growth parameters in multi-model inference analyses is recommended.