INFLUENCES OF AMOUNT OF PEDIGREE INFORMATION ON COMPUTING TIME AND OFMODEL ASSUMPTIONS ON RESTRICTED MAXIMUM-LIKELIHOOD-ESTIMATES OF POPULATION PARAMETERS IN SWISS BLACK-BROWN MOUNTAIN SHEEP
C. Hagger et M. Schneeberger, INFLUENCES OF AMOUNT OF PEDIGREE INFORMATION ON COMPUTING TIME AND OFMODEL ASSUMPTIONS ON RESTRICTED MAXIMUM-LIKELIHOOD-ESTIMATES OF POPULATION PARAMETERS IN SWISS BLACK-BROWN MOUNTAIN SHEEP, Journal of animal science, 73(8), 1995, pp. 2213-2219
Average daily gain between birth and 30 d of age of 42,644 lambs of Sw
iss Black-Brown Mountain Sheep were used in this analysis. The influen
ce of amount of pedigree information on computing time and on REML est
imates of population parameters was investigated on a subset of 7,848
lambs. If all available pedigree information was used, 89.4% of the la
mbs had at least four complete generations of known ancestors. For the
reduced pedigree information, only parents and grandparents of a lamb
were included. In the data set with complete pedigree information, 2,
616 additional animals (without records) caused 19.4% more equations,
4.6 times the number of non-zero elements in the system of equations,
and 21.2 times the computing time to reach convergence. The difference
in amount of pedigree information had only a marginal influence on th
e estimates of direct heritability (h(2)), maternal heritability (m(2)
), permanent environmental effects, and on the genetic correlation bet
ween direct and maternal effect (r(AM)) The complete data set of 42,64
4 recorded lambs was randomly split into four subsets to save computin
g time. In a fifth subset of 27,787 lambs (Set C) all combinations of
recorded grandparents and grand-offspring were accumulated because the
y contain information on the covariance between direct and maternal ef
fects (cov[AM]). Including cov(AM) in the model assumptions increased
estimates of h(2) and m(2) in all subsets. Estimates from Set C were s
mallest but showed the same trend. The estimate of r(AM) was always st
rongly negative, less than or equal to -.64. Taking flock-year contemp
orary groups (FY) random instead of fixed reduced h(2) to some extent
and increased estimates for r(AM) The estimates for h(2) and m(2) from
Set C without cov(AM) and FY fixed were .20 and .02, respectively. Th
e estimate for r(AM) from Set C