SOME FACTORS AFFECTING THE ESTIMATION OF GENETIC-PARAMETERS FOR CATTLE DYSTOCIA UNDER A THRESHOLD-MODEL

Genetic parameters for dystocia in the Main-Anjou breed were estimated. Data consisted of 28 178 birth records collected between 1978 and 1989 in 995 herds, with 161, 71 and 12 415 sires, maternal grandsires and dams, respectively, represented. Original scores (1 through 5) were collapsed in order to set two dystocia definitions: dystocia 1 (scores 1+2 v. 3+4+5) and dystocia 2 (scores 1 v. 2+3+4+5). Four models were proposed for genetic parameter estimation: (1) fixed effects plus sire effects; (2) model 1 plus maternal grandsire effect; (3) model 2 plus dam within maternal grandsire effects; (4) same as model 3 but a random effect 'herds' replaced a fixed effect 'regions'. Two methods of fitting models were applied: marginal maximum likelihood and the 'tilde-hat' approach. Estimates of genetic parameters by the two methods were similar. Models ignoring maternal effects overestimated the heritability of direct effects especially in the case of dystocia 2. Dystocia definition was responsible for the greatest difference among estimated genetic parameters. Possible reasons for this are discussed. When analysing large data sets, it is recommended judiciously to collapse dystocia categories and to apply approximate statistical procedures to complete models including maternal effects.