Genetic-parameter estimation of milk yield in White Maritza sheep breed using different test day models

The aims of this study were to estimate the genetic parameters of the test day milk yield (TDMY) of the White Maritza sheep breed population and to choose the most appropriate linear models for genetic-parameter estimation of test day milk yield. The White Maritza sheep breed is a multipurpose native sheep breed in Bulgaria. Test day milk yield data were collected from 1992 to 2015 (24 years). Milk yield recordings were made in 18 flocks according to the AC method (official milk recording by ICAR regulations). The database includes 8768 test day milk yield records belonging to 987 ewes. The pedigree file includes 1937 animals. Nine test day models (TDMs) were formulated and tested for the estimation of the genetic parameters of milk yield. The first three models were repeatability models (REP models), the second three were random regression models (RRMs), and the last three models were also random regression models with an added Ali and Schaeffer regression to describe the lactation curve using first-, second- and third-order polynomials. The average TDMY was 764.47 mL. There were no significant differences in the values of heritability ($$) calculated by the three REP models: REP1 0.355 $$ 0.060, REP2 0.344 $$ 0.047 and REP3 0.347 $$ 0.060. The same applied to the repeatability coefficients, which, for the three REP models, were 0.384 $$ 0.065, 0.376 $$ 0.051 and 0.378 $$ 0.065, respectively. Based on REP model 1, three models with random regression RRM1, RRM2 and RRM3 were constructed, which is associated with the use of first-, second- and third-order polynomials (for the random effects of both the animal and the permanent environment). The trajectories of $$ calculated by the three RRMs were not similar and demonstrated some differences, both at the beginning and in the middle of the milking period. The RRM with third-order polynomials demonstrated more genetic diversity until the 165th day of lactation, but Akaike information criterion (AIC), Bayesian information criterion (BIC) and log-likelihood (LogL) estimates were higher. The regression models with first- and second-degree polynomials were insufficient to reveal genetic diversity to a higher degree than REP model 1. The trend in the trajectories of $$ calculated by the three random regression models with Ali and Schaeffer regression models (ASRMs) was similar to that of random regression models without the Ali and Schaeffer regression incorporated. Although the noted advantages of the random regression models revealed, to a greater extent, the genetic diversity of test day milk yield, AIC, BIC and LogL estimates indicated that repeatability models achieved a better balance between complexity and fitness and a smaller prediction error compared to random regression models.