Reduction in accuracy of genomic prediction for ordered categorical data compared to continuous observations

Table 5 Estimates of heritability ( h ² )

Type of phenotype	Model	h ²	c	n = 1000	n = 2250
Continuous	Linear	0.10	-	0.16 ± 0.019	0.04 ± 0.004
		0.25	-	0.14 ± 0.018	0.16 ± 0.012
		0.50	-	0.39 ± 0.016	0.39 ± 0.011
Categorical	Threshold	0.10	2	0.56 ± 0.045	0.03 ± 0.003
			3	0.44 ± 0.042	0.03 ± 0.004
			4	0.37 ± 0.043	0.03 ± 0.004
		0.25	2	0.28 ± 0.039	0.10 ± 0.014
			3	0.27 ± 0.040	0.11 ± 0.013
			4	0.25 ± 0.037	0.11 ± 0.013
		0.50	2	0.62 ± 0.047	0.41 ± 0.018
			3	0.60 ± 0.045	0.39 ± 0.017
			4	0.62 ± 0.045	0.40 ± 0.017
Categorical	Linear	0.10	2	0.27 ± 0.037	0.01 ± 0.001
			3	0.26 ± 0.032	0.01 ± 0.001
			4	0.26 ± 0.031	0.01 ± 0.002
		0.25	2	0.09 ± 0.032	0.02 ± 0.004
			3	0.10 ± 0.030	0.03 ± 0.005
			4	0.11 ± 0.028	0.03 ± 0.006
		0.50	2	0.08 ± 0.009	0.10 ± 0.009
			3	0.10 ± 0.010	0.13 ± 0.010
			4	0.13 ± 0.011	0.15 ± 0.011

Estimates were obtained using Bayes C π linear model analysis of continuous phenotypes and Bayes C π linear and threshold model analyses of ordinal categorical phenotypes classified (c) as 2, 3 or 4 categories with heritabilities (h²) of 0.10, 0.25 or 0.50 in the Angus training (PT) population with (n) 1000 or 2250 observations.

ISSN: 1297-9686