The objective of this study was to compare animal model evaluations obtaine
d using test day models with different definitions of contemporary groups.
The data consisted of 170 937 test day records from 18 974 lactations of Bl
ack and White heifers calved in herds of small size. Seven test day animal
models which included contemporary group effect, fixed regressions on days
in milk, random animal, permanent environmental and residual effects were t
ested. The following contemporary group definitions were studied: herd-test
day (HTD), herd-year-season (HYS), herd production level month of the test
(HLMT), herd test day effects with clustered records from the smallest cla
sses, random HTD with fixed herd effect. Evaluations based on 305-day recor
ds were also obtained. The test day model with random herd-rest day effect
was the most efficient in describing milk yield variation. Models with HYS,
HLMT and clustered HTD were not as adequate as the model with fixed HTD ef
fect. The correlation between cows' lactation records (not included in eval
uations) with the mean predicted values of their parents showed further sup
eriority of the model that was based on random HTD ect. It was shown that t
he lactation model was not as good in predicting the future performance of
the animals as the TD model regardless of the definition of the CG group us
ed. Correlation between evaluations of bulls obtained, which were based on
TD model with random HTD effect and lactation model as well as TD model wit
h fixed HTD effect was the same, 0.96 whereas for the bet between-lactation
model and TD with fixed effect it was only 0.89 which means that introduct
ion of the TD model with random HTD effect into practice should be preferre
d.