TABLE 2

Reduction in average loss (PRIAL, in percent), for estimates of the genetic (), error (), and phenotypic () covariance matrix, together with mean entropy loss (×100) in unpenalized REML estimates of , the proportion of replicates (W, in percent) for which penalized estimation increased the loss in , and the mean tuning factor (ψ), for different constellations (A,…, K) of population values (see Table 1) and penalties on the original (ORG) or logarithmic (LOG) scale

ABCDEFGHIJK
125 families
23263612533148158279611555439
ORG
8663456541609483405544
7251291911326343111533
86321122011
W011332319028109
ψ949.8326.171.732.438.222.5948.8323.525.446.916.7
LOG
8862406955739685908187
7451261022136544171913
86312132111
W022132221061102
ψ521.869.510.62.38.92.0184.124.41.92.51.3
50 families
9010318338216945270479111021021868
ORG
9185807678709690476544
7565463427606453192663
97532122111
W0062511401758
ψ865.7411.584.934.035.720.0851.8407.626.455.013.6
LOG
9285777482819893888186
7666472739256555222817
87534232111
W00123021401152
ψ283.9121.620.95.37.92.4103.752.12.14.41.3
• Bondari's design with 125 or 50 families is shown.