This study aimed to estimate genetic parameters for total number of piglets born (NTLN), number of piglets born alive (NLNV) and number of piglets alive at five days of age (NLV5) using random regression models and to evaluate the best way for modelling the residual variance in the description of the genetic trajectories of litter size in Landrace and Large White breeds. The data came from a swine breed improvement program, and a total of 2388 and 2325 litter size records of Landrace and Large White, respectively were used in the analyses. The models considering homogenous residual variance showed the best goodness of fit for NTLN and NLV5 and the model with four classes of residual variances was most appropriate for NLNV (BIC). In the Landrace breed the maternal effect was not significant. The model including maternal effect and four classes of residual variance adequately described NTLN of Large White breed and the models without maternal effect and with homogeneous residual variance were most appropriate to describe NLNV and NLV5. The estimated heritability for NTLN, NLNV and NLV5 ranged from low to high (0.08-0.34, 0.04-0.29 and 0.05-0.21 in Landrace breed and 0.16-0.30, 0.10-0.37, 0.09-0.32 in Large White breed.). The magnitude of the rank correlations between breeding values of NLNV and NLV5 suggests that the recording of NLV5 is not necessary in this breed improvement program. High genetic gains can be obtained by selecting NLNV of primiparous females, due to the reduction in the generation interval.
piglet; correlation; heritability; model