Robust Inference for Variance Components Models in
Families Ascertained Through Probands: I.
Conditioning on Proband's Phenotype
Beaty TH, Liang KY
Genetic Epidemiology, 4(3):203-210 (1987)
Abstract
A robust approach for estimating standard errors of variance components by using
quantitative phenotypes from families ascertained through a proband with an extreme
phenotypic value is presented. Estimators that use the multivariate normal distribution as a
"working likelihood" are obtained by computing conditional ln-likelihoods, conditional first
and second derivatives in a Newton-Raphson approach. Robust estimates of standard
errors about the estimators are also provided. Tests of hypotheses are based on a
modification of the score test, which allows the assumption of multivariate normality to be
relaxed. Conditional goodness-of-fit statistics are proposed that can be used to examine
the fit of separate pedigrees to the overall model. This robust approach for estimating the
standard errors for variance components by conditioning on the proband's phenotype will
allow general inferences to be made from the analysis of families ascertained through
probands with extreme or unusual phenotypes and should be most appropriate for studying
many physiological traits that may be intrinsically nonnormal.
Variance component linkage analysis