A covariance structure model for quantitative genetic research.

Abstract

In this paper, we have presented a general covariance structure model for quantitative genetic research that incorporates measurement theory, components of variance, and regression theory for specification of structural relations among variance components. Each portion of the general model addresses specific issues important in delineating the genetic etiology of continuous traits. Through specification of a measurement model, genetic and environmental influences can be estimated independently of measurement error. Genetic and environmental sources of variance are defined from familial sampling designs through the components of variance model. Structural relationships among several traits at the genetic or environmental level can be specified through the regression model. The significance of parameters specified in each part of the model can be tested through model comparisons using a likelihood ratio chi-square test. We utilized the general model to test genetic covariance as a source of observed covariation of obesity and glucose tolerance in the Pima Indians. We initially applied the modeling approach using full-sib, half-sib sampling among the Pima with a second application including offspring as well as parental scores. Our initial applications of the methodology suggest that the covariance structure model can be a useful tool in genetic epidemiological research. Through model comparisons, our findings suggest that the association of glucose tolerance and obesity in the Pima is not due to a common set of genes but rather due to non-familial environmental influences. In conclusion, one task important for future research, as we see it, is the application of covariance structure models to other familial sampling designs and the evaluation of the usefulness of this approach through further applications to data.