The influence of a priori information on animal genetic evaluation was studied using a simulated 3.000 centimorgans genome, for a single trait governed by 800 loci, with two alleles for locus, and heritability varying from 0.40 to 0.60. Three populations with different sizes were derived from the base-population constituted by 1.500 males and 1.500 females: POP1 (100 animals), POP2 (300 animals) and POP3 (1.600 animals). Three a priori information levels were considered: no-informative priors (PNI), slightly informative priors (PPI) and informative priors (PI). The effect of using overestimated priors (with 50% error), considering slightly informative and informative priors on genetic evaluation was also studied. The influence of different levels of prior information on genetic evaluation was evaluated by the following criteria: error percentage related to the true value of variance components, Spearman correlation and Mean Square Error between real and predicted genetic values. POP1 and POP2 presented larger Error Percentages and were more sensitive to prior information. Larger levels of information led to best variance estimates and QME became smaller by increasing the information level, mainly for small populations. However, Spearman correlation were lower than 0.70 for POP1 and POP2, indicating significant changes in ranking of animals. Therefore, genetic evaluation of small populations is problematic, even if additional prior information is available.
a priori information; Bayesian analyses; genetic parameters; simulation; variance components
