Extended Pedigree Patterned Covariance
Matrix Mixed Models for Quantitative
Phenotype Analysis
Schork NJ
Department of Medicine, University of Michigan, Ann Arbor 48109-050
Genetic Epidemiology, 9(2):73-86 (1992)
Abstract
Overt computational constraints in the formation of mixed models for the
analysis of large extended-pedigree quantitative trait data which allow one
to reliably characterize and partition sources of variation resulting from a
variety sources have proven difficult to overcome. The present paper
suggests that by combining a restricted patterned covariance matrix
approach to modeling and partitioning the variation arising from polygenic
and environmental forces with an Elston-Stewart like algorithmic approach
to modeling variation resulting from a single genetic locus with large
phenotypic effects one can produce a model that is at once intuitively
appealing, efficient computationally, and reliable numerically. Extensions
and variations of this approach are also discussed, as are some simulation
and timing studies carried out in an effort to validate the accuracy and
computational efficiency of the proposed methodology.
Variance component linkage analysis