The predictive ability of indirect genetic models is reduced when culled animals are omitted from the data

Table 2 Summary statistics of validation datasets in the reference scenario full and in the five alternative scenarios

	full	noPhen	Testdays	noPhenGrp	noPhenGrp15	predPhen
Number of records	10,589	10,217	10,542	7489	10,105	10,589
Number of records on offspring of sires also represented in training data	4954	4354	4492	3208	4332	4954
Number of groups	928	927	928	663	885	928
Group size; mean (min–max)	11.9 (8–15)	11.5 (5–15)	11.8 (6–15)	11.8 (8–15)	11.9 (8–15)	11.9 (8–15)
% pigs without phenotypes	3.5	0	3.0	0	2.5	0
% groups including pigs without phenotypes	28.6	0	25.3	0	25.1	0
Additive genetic relationship within pens; mean (min–max)	0.13 (0.07–0.43)	0.13 (0.07–0.43)	0.13 (0.07–0.43)	0.13 (0.07–0.43)	0.13 (0.07–0.43)	0.13 (0.07–0.43)
ADG; mean g/d ± sd (min–max)	1049 ± 122 (411–1698)	1049 ± 139 (411–1698)	1049 ± 121 (411–1698)	1051 ± 128 (561–1698)	1050 ± 129 (411–1698)	1048 ± 120 (411–1698)

Alternative scenarios were: noPhen (individually culled pigs or pigs without phenotypes were omitted), testdays (pigs culled within the first 28% of test days were omitted), noPhenGrp (complete groups with culled pigs or pigs with missing phenotypes omitted), noPhenGrp15 (groups with more than 15% culled pigs or missing phenotypes were omitted), and predPhen (predicted breeding values assigned to pigs without phenotypes including culled pigs)

ISSN: 1297-9686