Estimates of variance components for test day records in an animal mod
el that considered multiple traits over multiple lactations were calcu
lated using REML methodology. Test day records were classified into II
periods within first and later lactations. Missing ancestors in the r
elationship matrix were classified in genetic groups. Data were collec
ted from Costa Rican dairy farms. Estimates of components for total an
d additive genetic variance were clearly heterogeneous during the lact
ation. Heritabilities for traits in later parities were slightly highe
r than those for traits in first parity. Heritabilities were highest f
or records of midlactation. Phenotypic and genetic correlations for ad
jacent test days were close to 1. Phenotypic correlations were lower t
han genetic correlations. Heterogeneity of variances during the lactat
ion suggests the adequacy of the multiple-trait test day model to desc
ribe milk yield during the lactation. When missing ancestors were allo
cated to a single base population instead of genetic groups, the estim
ates df residual variance were lower, and the estimates of genetic var
iance and genetic correlations were higher. When standardized records
were used instead of actual test day records, the estimates of residua
l and total variance were lower, and the estimates of genetic variance
were higher. Consequently, estimates of heritability and genetic corr
elations were also higher. Use of standardized data obtained by interp
olation procedures is not advised for estimation of genetic variance c
omponents in a test day model.