Table 2.

Cross-validation model fit statistics for six scenarios

Scenario	Training/prediction	Prediction ability (SE)	Slope (SE)	Mean top 10%
(a) Height
Random-CV	1558/413	0.58 (0.03)	0.73 (0.07)	0.38 (0.02)
Fullsib-CV	2024/39	0.22 (0.18)	0.39 (0.35)	0.19 (0.19)
4C	2063/451	0.29	0.39	0.29
4C Full-Sib	2063/57	0.16	0.16	0.33
4C Half-Sib	2063/29	0.23 (0.13)	0.23 (0.18)	0.17 (0.21)
4C Unrelated	2063/186	0.24	0.56	0.25
(b) Stem form
Random-CV	1558/413	0.78 (0.01)	0.81 (0.04)	–0.37 (0.04)
Fullsib-CV	2024/39	0.36 (0.16)	0.44 (0.22)	–0.09 (0.33)
4C	2063/451	0.57	0.71	–0.19
4C Full-Sib	2063/57	0.37	0.17	–0.02
4C Half-Sib	2063/29	0.29 (0.27)	0.26 (0.23)	0.07 (0.32)
4C Unrelated	2063/186	0.04	0.07	–0.11

For replicated scenarios, standard errors are provided after each estimate. Random-CV, random approximately fivefold cross validation within the ACE training population with 10 reps; Fullsib-CV, each full-sib family is predicted using a training set lacking any members from that full-sib family, with 51 reps; 4C, 451 fourth cycle trees are predicted using the ACE training population; 4C Full-Sib, fourth cycle trees within one family having full-sib relatives in the ACE training population; 4C Half-Sib, fourth cycle trees within seven families having half-sib relatives in the ACE training population; 4C Unrelated, fourth cycle trees not having direct parental relationships with the ACE training population.