Black Hereford Genetic Parameters and Predictions for Calf Traits with Alternate Modelling Strategies

Large breed databases allow use of multi-trait models for genetic parameters estimations (GE), genetic predictions (GR), and increased estimation accuracy. However, large databases are not always available to apply more complex models than single trait models. The objective of this study was to assess the potential of different variance structures (VS) to make GE and GR using multi-trait models with a small but rapidly growing database. Birth weight (BW), weaning weight (WW) and yearling weight (YW) records were obtained from the American Black Hereford Association. Single, bivariate and trivariate models were evaluated using a mixed model approach for each trait, and final models were selected through likelihood ratio test. Random components for BW (n=6821) and WW (n=5348) were direct genetic (DG) and maternal genetic (MG), covariance between direct and maternal additive genetic components and permanent environment (PE), while for YW (n=2667) only DG was included. The set of VS tested in the multi-trait models included results from the single model analyses, covariance fixations to zero, and definition or not of residual structure. Heritability estimates for DG ranged from 0.30 to 0.32 (BW), 0.19 to 0.27 (WW) and 0.28 to 0.31 (YW); for MG ranged from 0.07 to 0.11 (BW) and 0.02 to 0.03 (WW); for PE ranged from 0.05 to 0.06 (BW) and 0.09 to 0.16 (WW). EPD rank correlations between each pair of models for BW, WW, YW, birth weight maternal (BWM), milk (MK) and total maternal (TM) are presented in Table 1. These results show similar estimates to other breeds, and that, despite low record numbers, heritability estimates by different models were fairly similar by trait. EPD rank correlations had high values for almost every trait, where BW had the higher values and MK the lower and more inconsistent correlations. Finally, a correct definition of VS make possible to employ a trivariate analysis to assess traits with a low number of records without detrimental effects on estimations and accuracy of estimation.

Table 1. EPD rank correlations between model pairs.
Models/Traits	BW	BWM	WW	MK	YW	TM
1vs2	1.00	0.98	0.98	0.47		0.89
1vs3	0.99	0.94			0.93
1vs4			0.96	0.62	0.84	0.88
1vs5	1.00	0.90	0.96	0.47	0.84	0.91
2vs3	0.99	0.93
2vs4			0.95	0.33		0.80
2vs5	1.00	0.91	0.97	0.92		0.94
3vs4					0.81
3vs5	1.00	0.96			0.89
4vs5			0.99	0.52	0.96	0.94

Models nomenclature: 1, single model; 2, BW-WW; 3, BW-YW; 4, WW-YW; 5, BW-WW-YW.