Effects of number of records in each parity on accuracy of genetic parameter estimates for litter size at birth in pigs

The effects of unequal number of records of litter size at birth in each parity on estimates of variance components using REML estimation were assessed by a stochastic computer simulation. One male was randomly mated to ten females and each mating produced one male and two females as candidates for selection, assuming six generations including a base population without overlap. Two schemes were used: total number of records was constant (24,000 records) and total number of first-parity dams was constant (8,000 dams) through all generations. The number of records was assumed to be in the ratios 10:10, 10:5, 10:2 or 10:1 in the first and second parities or in the ratios 10:10:10, 10:5:5, 10:2:2 or 10:1:1 in the first three parities. The heritability and repeatability of litter size at birth were assumed to be 0.1 and 0.15, respectively. Mean litter size at birth was increased with a 0.5 increase in parity as a fixed effect. Two hundred replications were simulated for each combination of scheme and ratio of number of records in different parities. For each replicate, variance components were estimated, and the means and square root of their mean square errors (SMSE) were calculated. When the total number of records is constant, the number of parities, variation of number of records in each parity, and number of dams with records did not affect estimates of additive genetic variance and their SMSEs. Error variance estimates were always an upward bias of approximately 1%. However, the values of heritability and repeatability were slightly affected from error variance estimates because the amount of overestimate was small. SMSEs of permanent environmental variance and error variance increased with degree of unequal number of records in each parity. They were more accurate using the first three parity records than using first two parity records.