Read on Twitter
Have been going to do a thread on Accuracy (Acc) in relation to Australian Sheep Breeding Values (ASBV’s) for a while. Now seems like a good time.
(Sheep nerd alert! If you just follow for the pics, move right along
)
1/10 https://twitter.com/colinearl2/status/1361426484989108224
(Sheep nerd alert! If you just follow for the pics, move right along
)1/10 https://twitter.com/colinearl2/status/1361426484989108224
Let’s start with a thought challenge.
I have a sheep (random industry sheep) & I want you to tell me what the micron (mic) of its fleece will be. The only info you get is it’s a Merino
Now knowing it’s a Merino helps with your guess. It must be roughly between 12 to 26mic
2/10
I have a sheep (random industry sheep) & I want you to tell me what the micron (mic) of its fleece will be. The only info you get is it’s a Merino
Now knowing it’s a Merino helps with your guess. It must be roughly between 12 to 26mic
2/10
But Guessing randomly along this number line does not have the same probability of success. The distribution of mic fits a bell curve like this 
The guess with the highest rate of success will be the midpoint of 19mic, ....
3/10
The guess with the highest rate of success will be the midpoint of 19mic, ....
3/10
but the Acc of your guess would be 0 because the mic of my sheep could be anywhere along the number line.
But the important take away is that your guess is the midpoint (you did guess the midpoint, right?). My sheep’s mic has a 50% chance of above or below this point.
4/10
But the important take away is that your guess is the midpoint (you did guess the midpoint, right?). My sheep’s mic has a 50% chance of above or below this point.
4/10
Now if my sheep is a fine wool you can update your guess, 13 to 21mic, midpoint of 17mic. Your new guess 17mic, but this time you have some Acc (say 10%)
Knowing an animal’s parent’s mic the Acc of your guess improves again. The ends of the bell curve tighten with more info
5/10
Knowing an animal’s parent’s mic the Acc of your guess improves again. The ends of the bell curve tighten with more info
5/10
So, an ASBV for mic is the midpoint prediction of an animals genetic mic and the Acc is the shape of the bell curve.
What do these bell curves look like?
With some genetic parameters (mine are out of date) you can backwards calculate a bell curve from an Acc like this.
6/10
What do these bell curves look like?
With some genetic parameters (mine are out of date) you can backwards calculate a bell curve from an Acc like this.
6/10
Notice that the bell curve doesn’t change much from 20% Acc to 60% Acc. Going from 0% Acc to 99.9% Acc is not a straight line response. In fact you will never get to 100%!
Most people don’t mate just 1 ram, so the Acc of your team is important.
7/10
Most people don’t mate just 1 ram, so the Acc of your team is important.
7/10
Because an ASBV is the midpoint prediction each ASBV has a 50% chance of being to 
or too
The odds of 5 rams having an ASBV below prediction is 1 in 32, for 25 rams it’s 33.5million to 1!
The more rams in your team the more Acc your teams average will have. Like this
8/?
or too The odds of 5 rams having an ASBV below prediction is 1 in 32, for 25 rams it’s 33.5million to 1!
The more rams in your team the more Acc your teams average will have. Like this
8/?
NLW is a trait that will often have a low accuracy, (say about 20%)
Does this ASBV have value?
For 1 ram not much, but for a team of 25 rams yes (Team Acc 97%)
If you have 25 rams with a low Acc NLW of +10 you will still get a drop of daughters that perform at +10 for NLW
9/10
Does this ASBV have value?
For 1 ram not much, but for a team of 25 rams yes (Team Acc 97%)
If you have 25 rams with a low Acc NLW of +10 you will still get a drop of daughters that perform at +10 for NLW
9/10
Daughters from individual rams will vary a lot but your drop average will be very close to +10 for NLW.
(This all made sense to me, But I may have been to brief on some of the concepts. Just ask questions or ignore)
Cheers Mort
10/10
(This all made sense to me
, But I may have been to brief on some of the concepts. Just ask questions or ignore)Cheers Mort
10/10
