Methods of segregation analysis for animal breeding data: parameter estimates.

Three approximations to the combined model likelihood (including variation due to both polygenes and a major gene) used in segregation analysis were compared in their ability to estimate the parameters describing the model. Paternal half-sib data simulated under four different combined models were used. In general, when a combined model provides the best explanation of the data the parameter estimates from all three approximations investigated are, on average, in good agreement with the values simulated. The use of Hermite integration to replace the integration in the combined model likelihood provided the parameter estimates closest to those stimulated. The other two approximations, based on extensions of linear mixed-model theory, gave estimates further from the expected values, however, these latter two methods can estimate transmitting abilities for the sires directly from the analyses. The three approximations are similar in their ability to genotype sires at the major locus. On average, with the models studied here, about 65 per cent of sires can be successfully genotyped.