Over the past 5 years, genomic evaluations have been the hot topic for dairy cattle genetics. Yes, the technology is exciting. However, any new technology typically receives undue hype after its launch and unrealistic predictions of impact are promoted. The initial period of unwarranted hype about a new technology is usually followed by a time of reflection, which in turn is followed by periods of disillusionment and, eventually, proper implementation of the new technology.
Definitely, genomic evaluations of dairy cattle have been the topic of considerable hype and, perhaps, unrealistic expectations have been nurtured by some. Let’s face it – the potential positive impact of a new technology is likely to be emphasized and the potential downsides are likely to be glossed over. So, where are we on the hype/reflection/disillusionment/proper adoption path in regard to genomic evaluations? Probably, somewhere between reflection and disillusionment, but proper adoption will be the next step.
Reliability versus Accuracy
An overriding problem is interpretation of the “accuracy” of genomic evaluations. How is “accuracy” measured? Well, most would say by the reliability published alongside each genomic PTA (transmitting ability) for an individual trait or for an index that combines traits into a single value (such as Net Merit). However, are reliability and accuracy the same thing? No, they are not!
Accuracy has a specific statistical (scientific) meaning in addition to an intuitive meaning. Accuracy is a measure of the confidence range around a PTA – say, plus or minus 200 lb of milk or plus or minus $50 for Net Merit. Mathematically, accuracy for genetic evaluations is calculated as the square root of reliability. To put this into perspective, we will compare reliability versus accuracy alongside their corresponding confidence range for Net Merit for the four types of genetic evaluations of bulls – Parent Average (average of the PTA of his two parents), genomic test, first-crop daughters, fully proven (thousands of daughters in hundreds of herds). We will use a confidence range of 68% (which means there is about a two-thirds chance the eventual Net Merit of a bull will fall within the range).
The reliability of the genetic evaluations for Net Merit of most Holstein A.I. bulls is about 38% (Parent Average), 70% (genomic test), 85% (first-crop daughters), 99% (fully proven). These reliabilities suggest the result of a genomic test almost doubles the genetic knowledge known about a young bull without daughters. However, that is NOT the case. The accuracy of the genetic evaluations for Net Merit associated with those four levels of reliability are 62% (Parent Average), 84% (genomic test), 92% (first-crop daughters), and 99% (fully proven).