ebv debate
- Thread starter shearit by blades
- Start date
NZDan
Member
- Location
- Broomfield, New Zealand
The correlation between growth rate and adult weight is ~0.8 one of the stronger correlations.
http://www.sheepcrc.org.au/files/pa...ic_Parameters_for_Sheep_Production_traits.pdf
dogandstick
Member
- Location
- on the horizon
This was meant to the one that no one appears to have an answer for, but this muppet clicked on the wrong reply link
This was meant to the one that no one appears to have an answer for, but this muppet clicked on the wrong reply link
In statistical terms 90% accuracy is considered low. Statisticians prefer 3% or even 1% margin for error. A young ram when you buy it will never, ever, get anywhere near that.
NZDan
Member
- Location
- Broomfield, New Zealand
Different set of statistical laws.
Accuracy in ebvs is a measure of the possible change in ebvs up or down. Accuracy also improves in groups of animals, if your buying a group of sires the average ebv won't change much as for every animal that has a move another is likely to move the other way so the average stays similar.
How is that different from margin for error?
Isn't 75% accuracy is +/- 25% on the variance from mean?
So if a ram has an EBV of +2 for eye muscle but an accuracy of 50% he could add 1-3mm of muscle depth, but a ram with +1.5 but 99% accuracy will add 1.485-1.515mm (on the assumption that the lambs' dams are exactly average and all of the lambs from both sites are treated exactly the same)
NZDan
Member
- Location
- Broomfield, New Zealand
Thank you. That was very interesting.
Only 67% confidence in one standard deviation (only 96% confidence in two standard deviations!) and needing lots of linkages and/or hundreds of progeny to get to 99% accuracy.
To be frank, on that basis, if your buying by untested animals you may as well ignore EBVs altogether. It sort of proves the point that until everybody records it's a bit of a waste of time.
Only 67% confidence in one standard deviation (only 96% confidence in two standard deviations!) and needing lots of linkages and/or hundreds of progeny to get to 99% accuracy.
To be frank, on that basis, if your buying by untested animals you may as well ignore EBVs altogether. It sort of proves the point that until everybody records it's a bit of a waste of time.
NZDan
Member
- Location
- Broomfield, New Zealand
Not at all, remember buying one animal means the ebv may change. But if you buy two rams per year, the chance of one going up is equal to the chance of one going down, its important to remember that on average ebvs are accurate.
M-J-G
Member
What beef and probably sheep EBVs are missing in the UK is numbers, or the critical mass of information required, the system was not designed to run on such a tiny amount of input. The pedigree world is simply too tiny for the system to function to its potential. Which is a contributing factor in why this debate is 450 posts long and it hasn't convinced anyone of the success of EBVs. Which is odd when we consider that they are been on the horizon of being a big thing for a long long time now, yet it is still to be embraced.
To my mind there masses of information available from commercial cattle that could boost EBVs in both functionality and by reputation to the commercial farmer. As @neilo has pointed out sheep traceability is a long way behind that of cattle, so sheep may possibly not benifit from such a system. I do believe that the EBVs shouldn't end when a bull or ram is sold to a commercial farm, and that the commercial farm should be where his Ebvs are verified. For this reason I am glad to hear of projects like the beef efficiency scheme being set up, allowing commercial data to be collected from the coalface. Hopefully it will allow comparison across all breeds, and not just within breeds.
After all Ebvs are designed to benifit the commercial farmer, not the pedigree breeders.
To my mind there masses of information available from commercial cattle that could boost EBVs in both functionality and by reputation to the commercial farmer. As @neilo has pointed out sheep traceability is a long way behind that of cattle, so sheep may possibly not benifit from such a system. I do believe that the EBVs shouldn't end when a bull or ram is sold to a commercial farm, and that the commercial farm should be where his Ebvs are verified. For this reason I am glad to hear of projects like the beef efficiency scheme being set up, allowing commercial data to be collected from the coalface. Hopefully it will allow comparison across all breeds, and not just within breeds.
After all Ebvs are designed to benifit the commercial farmer, not the pedigree breeders.
M-J-G
Member
I don't disagree. Did anyone suggest an under supply?
dogandstick
Member
- Location
- on the horizon
And so it goes on.
Now then, if a ram with an index of say 300, which comprises the genetics of sire with, say 340, and a dam of 240, even with 90% accuracy, it's genetics is made up from so many variables with 16 great, great grandsires/dams, can there be any chance of estimating it's likely performance.
The only guarantee would be line bred animals perhaps.
Now then, if a ram with an index of say 300, which comprises the genetics of sire with, say 340, and a dam of 240, even with 90% accuracy, it's genetics is made up from so many variables with 16 great, great grandsires/dams, can there be any chance of estimating it's likely performance.
The only guarantee would be line bred animals perhaps.
- Location
- Mendips Somerset
,nearly all our main line sires are homebred and have been for 25 years , we use purchased new genetics on the sh ewes , and the best recorded sons used over the main flock within family groups ,for the next 4/5 years , its the only way you can magnify the best of your own genetics , and limit the influence of any problems the new ram WILL bring in , new sires are then sold within a few years unless they contribute nearly all good genes and very few poor ones .
nearly all our main line sires are homebred and have been for 25 years , we use purchased new genetics on the sh ewes , and the best recorded sons used over the main flock within family groups ,for the next 4/5 years , its the only way you can magnify the best of your own genetics , and limit the influence of any problems the new ram WILL bring in , new sires are then sold within a few years unless they contribute nearly all good genes and very few poor ones .
So it's a self fulfilling prophesy.
dogandstick
Member
- Location
- on the horizon
At the risk of patronising, I quote Wikipedia
"
"
A self-fulfilling prophecy is a prediction that directly or indirectly causes itself to become true, by the very terms of the prophecy itself, due to positive feedback between belief and behavior."
- Location
- Montgomeryshire
The ram you cite, would not have an accuracy of anything like 90% until there was more data on him than just sire & dam. He would start off at the midpoint between the two, if no performance data of his own was recorded, and then move towards his own performance as more data (& accuracy) was added. Similarly an unrecorded ram with no back pedigree recorded/input would start off with the average ebvs for his breed and the year of his birth (with very low accuracy) then move towards a more accurate estimate as data is added to the analysis.
When an animal gets closer to your 90% example, there would be a fair few related animals' performance linked into his analysis, each one making the estimate more likely to be accurate. Looking at our own breed's Stock Sire list, there are several rams that have 90% accuracy on their main ebvs, which have around 50 progeny recorded. There are shearling animals with 7 progeny recorded that are at 76% accuracy. For reference related lambs in well recorded flocks are generally in the low 70's until they get progeny recorded.
The ebvs provide an estimate of likely performance, which gets more accurate as more information is available, but it is still an estimate (the clues in the name). It is a valuable additional selection tool, to use alongside those of a stockman. I have never, and will never, advocate selecting an animal on paper estimates alone. That would be as daft as paying through the nose purely because someone said a lamb was by a ram that was worth 122,000 gns or whatever.