Bayesian methods for structural equation models in quantitative genetics with applications to the study of mammary gland disease

点击翻译按钮获取中文摘要

This project will develop quantitative genetic theory and statistical methods for studying complex genetic systems where recursive or feedback relationships between variables exist. Methods researched will be used to study mastitis, a disease of the mammary gland in breast-feeding women (mastitis), employing the cow as an animal model. Extensive health, milk production and milk somatic cell cow data from the Norwegian cattle health registry will be used to model relationships between variables. The 3-year project includes: 1) development of Bayesian Markov chain Monte Carlo algorithms for quantitative genetic systems under Gaussian assumptions. 2) Extension to systems in which some phenotypes are limited-dependent (e.g., binary). 3) Modeling of relationships between presence/absence of clinical mastitis, somatic cell concentration in milk, milk yield, genotype and several explanatory variables, using information from cow records. A fully pedigreed data set (complete medical treatment history of 33,453 first-lactation cows from 4961 herds, daughters of 245 sires) will be used for pilot studies.The research focuses on a new framework for the study of multivariate systems in quantitative genetics, using the cow as a model for a disease of the mammary gland, mastitis. Relationships between milk output, somatic cell concentration, presence or absence of clinical mastitis and several potential explanatory data will be investigated using data from the Norwegian cattle health registry system. Knowledge will be drawn and integrated from econometrics, structural equation modeling and statistical genetics. An educational component in a novel area in the interface between statistics, genetics, disease modeling, econometrics and sociometrics will be included. Software developed will be made available to the scientific community through the Internet.