The effect of errors in the economic weights on the accuracy of selection indexes
Genetics Selection Evolution
The effects of errors in the economic weights on the efficiency of index selection were investigated. A selection index for the genetic improvement of pigs was used as test case. Errors in single economic weights of ± 50 percent reduce the relative efficiency with less than i percent for all traits considered. Larger errors can result in considerable bias of the estimated genetic gain. The effect of errors in single economic weights are non linear and non symmetrical. Negative errors
... e errors (underestimation) are in general more critical than positive errors (overestimation). This dissymmetry depends on the scale used to define the errors. The effect of simultaneous sampling errors in the economic weight vector was studied by using both Monte-Carlo simulation and mathematical approximation. Result obtained by simulation indicate that the estimated genetic gain ( !H) and the realized genetic gain (toH i Î) are not normally distributed. For small sampling errors in the economic weight vector (coefficient of variation C.V < 0 . 50 ) the loss in relative efficiency was less than 2 .6 percent but increased to approximately i 5 percent for C.V. = i.o. .- The loss in relative efficiency of a selection index, the biases and variances of AH and OH ji T due to sampling errors in the economic weight vector are a function of the sampling variances and covariances of the economic weights and the genetic and phenotypic variance-covariance structure of the traits involved. AH is an overestimate of the maximum attainable genetic gain, while AH! I is an underestimate. The biases in OH and AH I are small for small sampling errors, but increase rapidly when the sampling errors in the economic weight vector increase. The variance of toR is relatively much larger than the variance of OH II, and increasing sampling errors in the economic weight vector make variance of JH increase at a much faster rate than the variance of toH! 1.