For this validation procedure, all the parameters of the models, including variance components, were re-estimated from TRN data in each of the TRN-TST partitions.
The Bayesian approach allows us to get easily standard error estimates of all parameters in the model, including variance components and their ratios.
This is a linear model that includes variance components to model the phenotypic similarities within a family, i.e. between the twins.
The model also included variance components, r0 representing variation associated with differences between individuals within practices, u00 representing variation associated with differences between practices and e representing random (measurement) error.
Linear mixed models include: estimation of variance (V), standard error (SE) and percentage of explained variance (%EV).
All models included a variance function to take into account variance dependence on fitted values.
The four weakly informative models include the conventional homogeneous variance model, the unrestricted heterogeneous variance model, the exchangeable variances model, and the consistency variances model.
Heterogeneity of variances was corrected using general least square (GLS) models including the 'varIdent' variance structure (different variance per factor level, this R-function is found in the 'nlme' package (Pinheiro et al. 2009).
Important classes of variance structure models include general linear models (GLMs [ 8] ) and structural equation models (SEMs [ 9]).
Different statistical models, including analysis of variance [ 46], linear regression [ 35], and their generalizations such as penalized regression [ 47], have been considered before.
