Estimates of components of variance and genetic parameters for reproductive traits by means of linear and threshold models

A total of 1883 repeated records of fertility (FERT), 1640 records for number of born lambs (NB L) and 1640 records for number of weaned lambs (NWL) from Santa Ines ewes were utilized to estimate (co) variance components and genetic parameters via linear (sire and animal) models and nonlinear threshold sire model. Models for FERT and NBL included the fixed effects of age and breeding weight of ewes. For both traits, herd-year-season, animal (ewe), or sire and daughter and residual were included as random effects. The model for NWL included the same effects, except that breeding weight was replaced by lambing weight. Restricted maximum likelihood-REML was used to estimate the (co) variance components for linear models, and for threshold model they were estimated using an approximation of the marginal maximum likelihood - MML, using the MTDFREML and CMMAT2 programs, respectively. Estimates of heritabilities obtained from threshold model were, substantially, greater than those obtained from linear models for all traits. Heritability obtained from linear sire model for FERT, NLB and NLW were 0.07, 0.07 and 0.04, respectively. In the same order these figures were 0.03, 0.09 and 0.01, when the animal model was applied. Estimates of heritabilities from threshold model were 0.12, 0.13 and 0.05 respectively. Estimates of repeatabilities were within the range of the literature and sometimes lower than heritabilities in this study, mainly when linear and threshold sire models were applied. Spearman rank correlation obtained between sire solutions using linear and threshold models was high and close to one.