Bibliographic citations
Vargas, A., (2017). Estimación de componentes de varianza utilizando los métodos bayesianos y máxima verosimilitud restringida para el estudio de la heredabilidad [Tesis, Universidad Nacional Agraria La Molina]. https://hdl.handle.net/20.500.12996/2889
Vargas, A., Estimación de componentes de varianza utilizando los métodos bayesianos y máxima verosimilitud restringida para el estudio de la heredabilidad [Tesis]. : Universidad Nacional Agraria La Molina; 2017. https://hdl.handle.net/20.500.12996/2889
@mastersthesis{renati/1116935,
title = "Estimación de componentes de varianza utilizando los métodos bayesianos y máxima verosimilitud restringida para el estudio de la heredabilidad",
author = "Vargas Paredes, Ana Cecilia",
publisher = "Universidad Nacional Agraria La Molina",
year = "2017"
}
The components of variance and heritability for milk production were estimated using a mixed linear model from the records of 3397 lactations from 1359 Holsteins cows in 57 herds with genealogical information of 5 generations using maximum restricted verisimilitude known as REML and Gibbs sampling based on Bayesian procedures. Both methodologies estimated moderate broad heritability of 0.135 via REML and average of 0.318 via Gibbs sampling. The exploratory analysis of residuals was done according to three types of residuals (marginal residual, conditional residual and random effects) of the mixed linear model estimated via REML. R functions were adapted to incorporate genealogical or pedigree information to the model. As a result of this, linearity of the fixed effects and normality of genetic component of the animal were verified. However, normality of random effects and conditional residuals were not found, neither homoscedasticity for the latter. In addition, it was found that the covariance structure considered in the model is not adequate for 132 animals. It was observed 215 animals and 7 herds with atypical effects. In the diagnosis of the Gibbs sampling simulation procedure from the Bayesian perspective, there were not problems of chain convergence, low Montecarlo errors and effective sample sizes greater than 1000 were obtained for each component of the model.
This item is licensed under a Creative Commons License