Analysis of live weight in Corriedale sheep: a distributional regression approach

Ossifo M.E.; Nakamura L.R.; Canaza-Cayo A.W.; Silva V.C.; Rosa L.F.; Ferreira D.F. · Scientia Agricola (2025) · DOI: 10.1590/1678-992x-2024-0219

Resumen

Understanding the growth and development of Corriedale sheep is critical to efficient production, as the breed is renowned for its high-quality wool and meat, as well as a variety of by-products. The main aim of this paper was to analyse the live weight of female Corriedale sheep according to five candidate variables: chest girth, abdominal girth, body length, wither height, and rump height. For this purpose, data from 100 animals from an experimental centre in Peru were examined. Generalized additive models for location, scale, and shape (GAMLSS), also known as distributional regression models, were used on account of their flexibility in describing the behavior of the response variable, allowing for the modeling of characteristics beyond the mean. Three different two-parameter distributions were tested to describe the response variable: gamma (GA), inverse gamma (IGAMMA), and inverse Gaussian (IG). A stepwise-based procedure was applied to select covariates for each parameter of the distributions. The GAMLSS based on the IG distribution was the most appropriate according to the Akaike Information Criterion (AIC), with the final model incorporating candidate variables in its two different regression structures (mean and dispersion). Finally, residual analysis indicated that the fitted model is suitable for describing the dataset under study.

Instituciones: Universidad Nacional del Altiplano