Table 5. MSE of the estimated variance components when the inverse covariance matrix was sparse.

Prior	5 QTL			50 QTL
	MSE ${\sigma }_e^2$	MSE ${\sigma }_a^2$	ρ	MSE ${\sigma }_e^2$	MSE ${\sigma }_a^2$	ρ
Uncorrelated P1	0.00092	0.00116	0.974	0.00133	0.00358	0.964
Correlated P2	0.01499	0.01413	0.858	0.00563	0.00465	0.911
Adaptive P3	0.00084	0.00110	0.969	0.00116	0.00390	0.960
Adaptive P4	0.00063	0.00070	0.965	0.00045	0.00266	0.951

Correlation (ρ) between the predicted and simulated genetic values; 10-fold cross-validation (size of training set $n=1000$). For comparison, results with P1 were taken from Table 3. ${\sigma }_e^2$, residual variance; ${\sigma }_a^2$, additive genetic variance.