Variance component estimation by resampling
- 12 January 1992
- journal article
- Published by Wiley in Journal of Animal Breeding and Genetics
- Vol. 109 (1-6) , 358-363
- https://doi.org/10.1111/j.1439-0388.1992.tb00415.x
Abstract
Summary: A method to avoid the inverse of the coefficient matrix of the mixed model equation is presented. This method makes use of Monte Carlo samples that have been simulated with the same information matrix as the data and is applied to approximate the restricted maximum likelihood variance component estimation via EM algorithm. The results were analyzed in a single trait and single record animal model. The different size data samples used to analyze this procedure were simulated with selection. The computational cost of this algorithm is proportional to the number of random variables in the model. The variability of estimates was satisfactory, and more efficient when sample size increased.Zusammenfassung: Varianzkomponentenschätzung mittels wiederholter StichprobenEine Methode zur Vermeidung der inversen Koeffizientenmatrix einer Misch‐Modell‐Gleichung wird vorgestellt. Sie verwendet Monte Carlo Stichproben, die mit derselben Informationsmatrix wie in dem Datenmaterial simuliert worden sind und wird zur Approximation der REML‐Varianzkomponentenschätzung über den EM Algorithmus angewendet. Die Resultate werden mit Ein‐Merkmal Eine‐Leistung Tiermodell analysiert. Die unterschiedliche Größe der Stichproben zur Analyse der Methode werden mit Selektion simuliert. Die Rechenkosten dieses Vorganges sind proportional zur Zahl der Zufallsvariablen des Modells. Die Variabilität der Schätzwerte war zufriedenstellend und effizienter bei Steigerung der Stichprobengröße.Resumen: Estimación de componentes de varianza mediante remuestreoSe presenta un método que evita invertir la matriz de coeficientes de las ecuaciones de modelo mixto. Este método utiliza muestras de Monte Carlo que han sido simuladas con la misma matriz de información que en los datos. Este método se utiliza para obtener una estimación aproximada de componentes de varianza por màxima verosimilitud restringida mediante el algoritmo EM. Se analizan los resultados en un modelo animal con un carácter y un registro por individuo. Las muestras de datos de diferente tamaño utilizadas para analizar este procedimiento se simulan con selección. El coste computacional del algoritmo es proporcional a la cantidad de variables aleatorias incluidas en el modelo. La variabilidad de las estimaciones obtenidas es satisfactoria, y más eficiente cuando el volumen de información aumenta.Keywords
This publication has 6 references indexed in Scilit:
- Restricted Maximum Likelihood Estimation of Variance Components in Animal Model Using Sparse Matrix Inversion and a SupercomputerJournal of Dairy Science, 1990
- Restricted maximum likelihood to estimate variance components for animal models with several random effects using a derivative-free algorithmGenetics Selection Evolution, 1989
- Four Equivalent Sets of Mixed-Model Equations with Relationship Matrix for Estimation of Genetic ParametersJournal of Animal Science, 1988
- Estimating Variance Components in a Class of Mixed Models by Restricted Maximum LikelihoodJournal of Dairy Science, 1986
- Comparison of an Animal Model and an Equivalent Reduced Animal Model for Computational Efficiency Using Mixed Model MethodologyJournal of Animal Science, 1984
- Computing the Diagonal Elements and Inverse of a Large Numerator Relationship MatrixPublished by JSTOR ,1976