Manipulating response to selection on many traits

While selection index theory is a solid framework for a linear optimization of multiple trait selection, it does not always provide desirable selection responses. Responses can be manipulated by varying the economic weights used in a multiple trait index. However, trait responses depend not only on weights, but also on accuracies of the estimated breeding values (EBVs) as well correlations between these EBVs. Predicted selection responses for some traits can be in an undesirable direction if the traits have unfavourable correlations and if there is a large difference in the economic values of the standard deviation of the EBV, as in that case one trait dominates over the other. We extend insight in a two-trait example to a more general case with many traits in the breeding objective. We propose to investigate for each trait its correlation and its dominance relative to the other traits. These parameters can help determine a strategy to avoid undesirable trait responses, whether by changing economic weights or by aiming for more accurate EBVs, or by a combination of these.