Table 6.

Prediction accuracies for wood quality and growth traits estimated with cross-validation sets built with individuals within full-sib families for wood quality and growth traits using ridge regression (RR) for subsets (SS) of SNPs from single-marker regression controlling for A or K , the largest absolute effects from the full model (RRA), the highest minor allele frequencies (HMAF), and a random selection (RAN) for various subpopulations of individuals

	Accuracies obtained in cross-validation with various marker subsets
Trait†	Number of markers (m) and subpopulation*	Association tests using pedigree information(CVSSA)	Association tests using genotypic information₸(CVSSK)	Markers with largest absolute effects (CVSSRRA)	Markers chosen randomly (CVSSRAN)	Markers with highest minor allele frequencies (CVSSHMAF)
ADEN	m	640	404	600	600	600
	BG1	0.76	0.78	0.74	0.78	0.78
	BG1 → BG2	0.20	0.06	0.03	0.06	0.16
	S1	0.74	0.77	0.74	0.71	0.73
	S1 → S2	0.55	0.51	0.52	0.58	0.61
AMFA	m	406	359	600	600	600
	BG1	0.74	0.78	0.72	0.75	0.74
	BG1 → BG2	0.03	0.04	0.01	−0.12	0.11
	S1	0.73	0.76	0.69	0.65	0.66
	S1 → S2	0.51	0.54	0.44	0.55	0.52
HT17	m	476	426	600	600	600
	BG1	0.64	0.70	0.69	0.56	0.57
	BG1 → BG2	−0.03	−0.04	−0.07	0.00	−0.19
	S1	0.72	0.77	0.74	0.63	0.64
	S1 → S2	0.28	0.29	0.29	0.32	0.29
DBH17	m	482	403	600	600	600
	BG1	0.56	0.61	0.58	0.49	0.51
	BG1 → BG2	−0.12	−0.13	−0.25	−0.05	−0.23
	S1	0.71	0.77	0.70	0.63	0.65
	S1 → S2	0.17	0.18	0.18	0.22	0.20
MSE‡		0.025	0.024	0.023	0.024	0.023

Subset of significant SNPs (P < 0.05) identified after an association study using relatedness estimated with pedigree information.

^₸Subset of significant SNPs (P < 0.05) identified after an association study using the realized genomic relationship matrix.

^†ADEN: average wood density; AMFA: average microfibril angle; HT17: 17-year height; DBH17: 17-year diameter at breast height.

*BG1: both training and testing sets from breeding group 1; BG1 → BG2: training sets from breeding group 1 and testing sets from breeding group 2; S1: both training and testing sets from site 1; S1 → S2: training sets from site 1 and testing sets from site 2.

^‡Mean standard error over cross-validation iterations and subpopulations.