Fitting of fixed regression curves with different residual variance structures for Nellore cattle growth modeling.
Different polynomial functions were tested for mean trajectory modeling with different residual variance structures. A total of 15,148 weight records of 3,115 Nellore Mocho cattle with ages between 1 and 660 days, raised in northern Brazil. First, the mean trajectory of cattle growth curve was fitted by a fixed regression using orthogonal polynomials with orders ranging from two to seven. Analyses were performed using the least-squares method, disregarding animal and/ or maternal random effects. Then, the best model was evaluated using different residual variance structures and homogeneous and heterogeneous classes. We considered as fixed effects those of groups of contemporary and of dam age at birth (as linear and quadratic covariate). The random model part included animal and maternal effects (direct genetic and permanent environments). We concluded that the estimates of variance components and genetic parameters were affected by both fixed regression curve polynomial order and residual variance structure. Moreover, random regression model considering an order-four polynomial function with a fixed curve and six-class residual variance showed better fits.
Main Authors: | , , , , , , |
---|---|
Other Authors: | |
Format: | Artigo de periódico biblioteca |
Language: | Ingles English |
Published: |
2020-06-23
|
Subjects: | Curva média, Modelagem residual, Modelo linear, Parâmetro Genético, Regressão Linear, Linear models, |
Online Access: | http://www.alice.cnptia.embrapa.br/alice/handle/doc/1123435 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | Different polynomial functions were tested for mean trajectory modeling with different residual variance structures. A total of 15,148 weight records of 3,115 Nellore Mocho cattle with ages between 1 and 660 days, raised in northern Brazil. First, the mean trajectory of cattle growth curve was fitted by a fixed regression using orthogonal polynomials with orders ranging from two to seven. Analyses were performed using the least-squares method, disregarding animal and/ or maternal random effects. Then, the best model was evaluated using different residual variance structures and homogeneous and heterogeneous classes. We considered as fixed effects those of groups of contemporary and of dam age at birth (as linear and quadratic covariate). The random model part included animal and maternal effects (direct genetic and permanent environments). We concluded that the estimates of variance components and genetic parameters were affected by both fixed regression curve polynomial order and residual variance structure. Moreover, random regression model considering an order-four polynomial function with a fixed curve and six-class residual variance showed better fits. |
---|