Archive for Best Linear Unbiased Prediction

Genetic Accuracy – Can you trust the numbers?

Dairy breeders are continually taking steps to be more exact about the way they farm and the products they buy, produce and sell.  However when it comes to the genetic make-up of our animals there remains significant difference of thought, amongst breeders, about the actual accuracy of the genetic information. Breeders are presented with a wide range of facts. Gold Medal, Extra, Star Brood, DOM, proven, genomic, photos, Supreme Champion … no wonder many breeders are confused. The Bullvine feels that breeders need to be objective about the animal information they see and to think in terms of the accuracy of the information. Now we are not talking about whether or not an animal meets the ideal. We’re talking about how much we can rely on the facts we see in hard copy or from virtual communication sources.

In the Beginning

In the nineteenth century milk cows were mostly dual purpose and herd size was small. People wanting to get into dairying purchased a cow or bull based on what the seller said were the animal’s merits. In time breed societies were formed to document lineage. That was followed in the early twentieth century with third party authentication of both yield and conformation.  The third party oversight of parentage and performance were the beginning steps to know the accuracy of the information. That was the start.

Many Steps Along The Way

Having a milk record or type classification authenticated for a single one cow in a herd was initially thought to be very useful information. The next move was to compare a cow to her dam to see if improvement had been made.  But that did not help much as the cow and her dam were not simultaneously at the same age and, in some cases, not in the same herd. Of course, over time we have learned that we need to know the performance of the cow’s herdmates. That was the stage where breeders started to compare animals within a herd with the desire to know which animals were superior, or, conversely, inferior for a trait. The biggest breakthrough in accurately determining the relative genetic merit of an animal came when Dr Charles Henderson, Cornell University, developed the analysis technique that he called B.L.U.P. (Best Linear Unbiased Prediction).  Forget about trying to understand the term, what it does is compare all animals within a herd and then compile the results across all herds to produce genetic rankings for males and females.

What About Accuracy?

From a genetic merit perspective it is important to know two things. Firstly where does the animal rank in the populations? And secondly, and also very important, how accurate is the prediction? How much trust can a breeder put in the animal’s genetic rating? If information is of limited accuracy, then it may be nice to know, but it does little for constructive breeding or to provide the opportunity to drive up on-farm profits. Accuracy produces confidence; confidence accelerates advancement, and negligence ruins the reputation which accuracy had raised. (Read more: Has Genomics Knocked Out Hot House Herds? And The Hot House Effect on Sire Sampling)

Let’s Compare Accuracy

The range in accuracy of genetic evaluation indexes goes from 0 to 99% and is called Reliability. The following chart is an approximation of the accuracy of predicting an animal’s total merit index (i.e TPI, NM$, LPI, or any other national total merit index) from the information that is known on the animal.

Reliability In Predicting An Animal Total Merit Index

Genetic Accuracy – Can you trust the numbers2

As far as accuracy goes the winners, as a result of incorporating genomic information into our genetic evaluation systems, have been young bulls, young heifers and brood cows. Adding genomic information has resulted in a doubling of the accuracy of their indexes. For further information on accuracy an interesting read is Two Ways to Look at Accuracy for Genomic Young Bulls published by Canadian Dairy Network.

What’s Ahead?

As more and more animals are genomically tested and recorded for their performance, the accuracy of all genetic indexes will increase.  Three other steps that will assist in increasing the accuracy of total merit indexing are needed:

  1. Have every milk weight, fat %, protein% and SCC automatically captured at every milking;
  2. Have information on new economically important traits collected;  and
  3. Have more economic information available on more traits.

Breeders will be the benefactors of having more and more accurate information so that they can make more and more accurate decisions.

The Bullvine Bottom Line

Having genomic information has been a significant step forward for increasing the accuracy of genetic indexes. But it will go beyond genetics and genomics in the future. Read past Bullvine articles for further details about genomics for health and management (Read more: Herd Health, Management, Genetics and Pilot Projects: A Closer Look at ZOETIS) and what lies beyond genomics (Read more: Forget Genomics – Epigenomics & Nutrigenomics are the Future). When buying genetics breeders need to check that the animals, semen or embryos they are considering will both follow their breeding plans (Read more: What’s the plan?) and that the information is accurate.  Breeding dairy cattle is faster paced every year. The accuracy of the information used is an important consideration.



Get original “Bullvine” content sent straight to your email inbox for free.





Has Genomics Knocked Out Hot House Herds?

For years there has been an unspoken awareness that some herds appeared to be able to “work the genetic indexing system.”  These herds clearly understood how genetic indexing systems work in their country and how to manipulate the composition of their herd in order to achieve the highest ranking possible for some top members of the herd.  It was possible to pick these herds out, they had top females and even though they had many sons sampled out of them, seldom were able to produce a top ranking sire, especially for total merit.  This scenario played out in all major countries that use the BLUP (Best Linear Unbiased Prediction) system.  Fast forward to the top total merit rankings since genomic testing has become available.  At first glance it appears there might be a lot less of those animals on the list.  The question is: “Has genomics knocked out all of these animals from the top lists?”

How It Was Done In The Past

First it is necessary to understand what was happening in the past.  Since indexing calculations are based on the variance compared to herd and genetic base, the greater the difference the greater the gain or loss in the result.  Putting that into practice takes two areas: conformation and production evaluations.  The following is a breakdown on how both were done.


In order to get maximum results from their genetics programs, many top programs needed to have their top cattle score significantly higher than the other animals scored that same day.  While many people deemed these herds “Hot Houses”, in reality they are just working the BLUP system to get maximum results.  Since the calculations also took into account the genetics of the other animals scored, these “hot house” herds needed to have daughters of high type bulls that would score lower than the selected cattle that were typically sired by bulls with lesser conformation scores.  For example, you have a low value cow sired by a +14 conformation sire that goes 79 points, and a high conformation cow sired by a +6 conformation sire that goes 86 points.  This would provide the selected cow with the greatest difference over the expected value and have significant improvement in their EBV for conformation and thereby in their overall total merit.

It’s in these herds that you may have seen a “good group” and a “bad group” of the herd, with a corresponding difference in management and presentation of the groups.  While it’s normal in any herd to have the high value or “family favorites” get some level of preferential treatment, these herds took it to a new level.  While this sounds bad, in reality it was necessary in order to achieve top rankings.  For the classifier visit the good group was show ready and the bad group was ready to head for beef.  (Though if you read “Tom Byers: It’s Classified” you realize that this really does not make that big a difference for the professional classifier).

In contrast to the “hot house” herds who try to have a high herd average score (for example the average 2yr old score of 83+ points) find it very hard to get high indexing conformation females.  With very little difference in scores from the top to bottom of the herd, there is less herd variance, contributing to a lowering of their overall rankings.  Since these herds where not a cross section representation of the breed population and BLUP treats them as if there were, these cattle actually get somewhat penalized for being a member of a great herd.

In order to have maximum impact, herds wanting to have high index’s needed to have maximum within herd variance. This meant that they have to have a true cross section of the breed present in their herd, as opposed to just the best of the best, like many of these breeders would have liked. It’s also for these reasons why niche type sire sampling programs need to be used in all types of herds not just high conformation breeding programs.


The story is not that different on the production side.  Here the comparisons are for milk, fat and protein yields on a within test day basis.  Adjustments are made for a cow’s age, lactation number, stage of lactation, month / season…etc.

In order to maximize the increase in production genetic evaluations, these “hot house” herds needed to  have underperforming  daughters of high production sires, that were being out produced by the selected females that were typically sired by more balanced sire who’s production index may not be as high.  In Canada, this is where you would see females with very high (i.e. +200 and more) BCA deviations.  Sometimes you would see deviations that were greater than even their herd average BCA.  You ask yourself “How could one cow on the same feed, same treatment, same exact program, produce twice as much milk as another cow?” While it sounds unrealistic, it was necessary in order to gain maximum results.  All breeders have seen cows that can out produce herd mates by 30, 40 even 50%, but when you see them doing more than double (100%) it raises questions in the minds of people with practical cow sense.  Hence why some herds are stamped as “Hot House’s.”

How Genomics Has Changed Things

relative weighting for Direct Genomic Value (DGV) compared to traditional Estimated Breeding Values (EBV)With the introduction of genomic evaluations in August 2009, the effects that any “hot house” efforts can have has been reduced in the genetic indexing systems.  This is because for young cows in first or second lactation, the relative weighting for Direct Genomic Value (DGV) compared to traditional Estimated Breeding Values (EBV) is roughly 55:45 (Source: Canadian Dairy Network).  What that means is that if a “hot house” cow would have had a 300 point jump from these types of efforts, they now would only see a 165 point jump.  While it would still have an effect, genomics has greatly decreased the “hot house” effect. Remember that the female family members of each cow are being re-evaluated as well.  Additionally those females formerly lower on the listings, but that were in herds where practices are normal, could now move up the genetic index rankings.

The other factor that Genomics has brought into play is that, if a particular animal is not gifted with the best genomics her parents had to offer, she will also see a significant drop.  So let’s say that a cow has an EBV-PA of +2500 LPI or TPI, but her genomic panel comes back with a LPI or TPI value of +1500.  That cow would see a drop of about 450 points.  Dropping her  to an LPI or TPI of +2050. This takes her  from being near the top of the list to almost out of consideration.  All this is outside the control of any on-farm practices.  It’s for these reasons I am sure that some owners now get nervous when opening their genomic results letters.  This single test can have the biggest effect on the genetic profitability of any cow.  It can even have a greater effect than the classification.  With GLPI’s and GTPI’s now over 60% reliable, adding animal performance information now has much less influence than in the past.

The Bullvine Bottom Line

The great news is that genetic indexes that contain animal genomic information are not as influenced by preferential treatment or herd variance as traditional genetic indexes are.  Since genomic values are based on evaluations of thousands of cattle in many different herds, in many different environments, and in different countries, the ability of a “hot house” to greatly change results has been significantly diminished.  That is not to say it has been totally removed.  Remember that 45% of the new GLPI formula is still based on an animal’s performance compared to contemporaries.  Therefore, these efforts will still have an effect.  It is for these reasons that you see some previously prominent cows and cow families are now absent from the top female lists.  Am I saying that these cattle may not be great investment?  No, what I am saying is consider these factors when making your purchase decision.  Do your homework before selecting, breeding, merchandising or buying.  GLPI’s, GTPI’s and DVG’s will help you make more informed decisions, but remember they are just a tool.
The Dairy Breeders No BS Guide to Genomics


Not sure what all this hype about genomics is all about?

Want to learn what it is and what it means to your breeding program?

Download this free guide.

Send this to a friend