We
present the final model using REML and checked for the absence of
significant correlation among independent variables and for
normality of residuals. To assess the overall goodness of fit we
provide the correlation between the fitted and the observed values
(Bogner et al., 2010). We also calculated Edwards’ R2 statistic
(Edwards et al., 2008) for the fixed effects, which compares the full
model with a null model with all fixed effects deleted (except
typically the intercept) while retaining exactly the same covariance
structure. Though it is conceptually different to the partial R2 in
linear regression, this statistic can be similarly interpreted in the
sense that it measures the marginal improvement or reduction in
unexplained variability in the fixed component after accounting for
a given predictor effect.