Table 2.

Number of eigenvalues explaining 90, 95 or 98% of the variance for genomic relationship matrices

Option	Number of eigenvalues
	100 QTN			1000 QTN
	90% eigenvalue	95% eigenvalue	98% eigenvalue	90% eigenvalue	95% eigenvalue	98% eigenvalue
60 k	8496	12,185	16,978	8502	12,192	16,984
60 K-BL5	9553	13,787	19,111	9560	13,796	19,120
60 K-GWAS3	4571	7537	13,139	4757	7704	13,230
60 K-QTN-BL5	9553	13,788	19,112	9563	13,806	19,136
60 k-QTN-BL5-TRUEd	76	1803	5093	469	1942	5140
60 k-QTN10-BL5-TRUEa,b,d	4054	8972	15,886	7482	13,320	19,918
60 K-QTN-BL5-GWAS3	4082	7084	12,880	4627	7594	13,186
QTN	88	94	98	793	872	930
QTN-BL5c	94	122	7639	863	980	7925
QTN-BL1c	89	95	127	806	888	995

Options used to construct the genomic relation matrix: 60 k non-coding SNPs (60 k), all causative QTN (QTN), the top 10% causative SNPs (QTN10), blending at 5% (BL5) or 1% (BL1), weighted by the 3rd iteration of the single-step GWAS (GWAS3), and weighted by true QTN effects (TRUE) for datasets with 100 or 1000 causative QTN

^a10 eigenvalues explained 76% of the variance of $\mathbf{G}$ for the 100-QTN scenario

^b100 eigenvalues explained 71% of the variance of $\mathbf{G}$

^cEigenvalues after number of QTN (100 or 1000) had values approaching 0 (below 10E−4)

^dSimulated true weights for QTN and a constant equal to the minimum QTN value for SNPs