Íslenskar landbúnaðarrannsóknir - 01.03.1979, Qupperneq 92

90 ÍSLENZKAR LANDBÚNAÐARRANNSÓKNIR
material used in this study. The parents,
at least the stallions are most certainly
supposed to be selected on their own per-
formance or the performance of close rel-
atives. The measured offspring are also
almost certainly above average of the
population, as the total performance or
point scoring index is concerned. On the
other hand the total performance is distri-
buted on the ten traits, which must reduce
the actual selection pressure on each trait
considerably.
Both selection on parents and offspring
records lead to a proportionally greater
reduction of the genetic variance than of
the phenotypic variance between the sel-
ected animals. When the heritability is
estimated from the intraclass correlation
of half-sib, both selection on parents and
offspring result in reduced heritability
estimates (Cochran, 1951; Ronningen,
1972a). Heritability estimated by reg-
ression ofoffspring on parent is not biased
by selection on that parent side (RöNNIN-
GEN, 1972b) On the other hand such sel-
ection causes high standard errors on the
heritability estimates and makes them
sensitive to any bias introduced into the
numerator of the regression due to reduc-
ed variance of the independent variate,
which makes up the denominator.
The material used in this study offered
possibilities of additional estimates on
genetic parameters from parent-offspring
relationship, as measurements on the par-
ents were available in many instances.
Due to certain confoundings (Hill and
Nicholas, 1974) between the estimates of
genetic parameters, obtained by different
methods of estimation from the same mat-
erial, this was not thought to be worth-
while in this study. A theoretical study on
a maximum likelihood procedure of solv-
ing such problems seems promising (Hill
and Nicholas, 1974: Hill, 1975:
Thompson, 1976), but has not found
practical application so far.
The second item, already mentioned,
which is likely to introduce bias into the
estimates of genetic parameters, is the
possibility of environmental covariance
adding to the observed similarity of the
relatives used for the estimation, in this
case the half-sibs. This kind of bias is lik-
ely to have inflated heritability estimates
of racing ability in several horse populat-
ions (Cunningham, 1976; Kieffer,
1976). The presence or amount of such
correlation in this material is not so easily
predictable. As the bias caused by the sel-
ection and the environmental correlation
between half-sibs should point in opposite
directions, the best one could hope for is
that these two kinds of biases counteract
each other.
A non-random mating within the pop-
ulation is not likely to cause marked errors
in the estimates of the genetic parameters.
The matings tend to be disassortative as
well as assortative, in terms of individual
traits, although assortative mating is pro-
bably more common for the total score.
The separate run determined to estim-
ate the genetic parameters for the traits in
the female population yielded heritability
estimates of similar magnitude as those
listed in table 5. No significant difference
was observed but slightly higher herit-
ability estimate of „legs“ and slightly low-
er heritability estimates of the gaits were
obtained from the records on females. As
the mares make up a great part of the
whole material used for the previous est-
imation a significant difference was hardly