An Extreme Value Regression Approach to the Prediction of Shrimp Size Distribution

Abstract

A statistical non-normal regression model was developed to characterize the growth of white shrimp (Litopenaeusvannamei) and the shrimp’s weight distributions throughout the fattening process. Empirical data was collected from submersible cages in the Gulf of California, Mexico. In this article, the authors demonstrate the efficiency in predicting the shrimp size distribution by using the extreme value regression methodology. The extreme value regression has been used in quality control engineering, reliability and survival analysis; however, it has yet to be applied in aquaculture setting. Findings suggestthat the extreme value regression was the best model to fit the weight of shrimp. The extreme value regression model can be used to predict not only the average weight but also the shrimp size distribution and the percentiles as a function of the number of days that the shrimp stay in the farm.