Leaking genetic gain: not quite a selection index


A few years ago I was talking with a breeder who spent substantial time running genetic analyses, with fairly sophisticated linear mixed models. From there the breeder obtained BLUPs for a couple of selection criteria (using an animal model with heterogeneous site variances), heritabilities, genetic correlations, etc. The typical stuff.

Then came the leak: let’s say that the criteria were stem diameter (dbh) and wood basic density (den). The breeder thought that dbh was twice as important as den, therefore used a selection index weighing the breeding values as I = 2 * dbh + 1 * den. Easy.

My jaw dropped, because I couldn’t see how the breeder was accounting for the different genetic variances, the heritabilities, the genetic correlation or the relative economic importance of the criteria. I couldn’t because the breeder was not considering any of that.

The breeding programme used a “faux selection index”. At first sight it looked the part, but it completely wasted all the effort used in the genetic analyses. Why? Someone was trying to be clever and nobody else audited the methodology.

, ,