Skip to main content

Table 2 Set of 16 statistics used to compare predictions based on the whole and partial beef cattle datasets

From: Semi-parametric estimates of population accuracy and bias of predictions of breeding values and future phenotypes using the LR method

Statistic Description
\(h^{2}\) REML estimate of heritability for each ‘Partial’ dataset (each random 50% missing)
\(b_{w,p}\) Regression of whole on partial EBV (expectation of 1.0)
\(b_{w,p}^{r}\) \(b_{w,p}\) computed within reference samples (i.e. Those with phenotypes maintained in the creation of the partial sample)
\(b_{w,p}^{v}\) \(b_{w,p}\) computed within validation samples (i.e. Those with phenotypes treated as missing in the creation of the partial sample)
\(b_{p,w}\) Regression of partial on whole EBV (expectation depends on accuracies)
\(b_{p,w}^{r}\) \(b_{p,w}\) computed within reference samples
\(b_{p,w}^{v}\) \(b_{p,w}\) computed within validation samples
\(\rho_{w,p}\) Correlation between whole and partial EBV (expectation depends on accuracies)
\(\rho_{w,p}^{r}\) \(\rho_{w,p}\) computed within reference samples
\(\rho_{w,p}^{v}\) \(\rho_{w,p}\) computed within validation samples
\(r\left( {y_{r} ,\hat{u}_{r} } \right)\) Correlation between the partial EBV and the adjusted phenotypes for the reference samples
\(r\left( {y_{v} ,\hat{u}_{v} } \right)\) Correlation between the partial EBV and the adjusted phenotypes for the validation samples (NB. This is the conventional measure of accuracy in cross-validation genomic selection studies)
\(d_{w,p}^{r}\) Difference between whole and partial EBV (in absolute value) computed within reference samples
\(d_{w,p}^{v}\) Difference between whole and partial EBV (in absolute value) computed within validation samples
\(Vd_{w,p}^{r}\) Variance of the difference between whole and partial EBV computed within reference samples
\(Vd_{w,p}^{v}\) Variance of the difference between whole and partial EBV computed within validation samples