|
Post by shumakerherefords on Oct 26, 2013 20:11:51 GMT -6
Tom, I really don't see how checking the milk volume on a few cows would allow them to estimate milk production across all blood lines in the Hereford breed. Bet it was an interesting project though, but with Kstate facilities that eliminated a lot of the danger!
|
|
|
Post by strojanherefords on Oct 27, 2013 12:25:06 GMT -6
I think that the idea behind EPD's is a good one, however since we are on a breed wide program we are ultimately dependent on each other to properly administer TPR. To illustrate the complexity administering the record keeping consider this not so abstract thought experiment... My ideal weight for a mature cow is 1300-1400 pounds and I plan on accomplishing that goal by sorting the cows by weight into three groups: cows under 1300 pounds, cows 1300-1400 pounds and cows over 1400 pounds; and breeding those cows to two bulls. I will breed the cows under 1300 pounds to bull A to increase frame size, and I will breed the cows over 1400 pounds to bull B to decrease frame size. The rest of the cows, I do not care what bull they are breed to so I will flip a coin to pick the bull to use. These cows will be managed identically from breeding to weaning. So the question is how many comparison groups are there? 2, heifer calves and bull calves. If you castrate some of the bull calves prior to weaning, then those calves should be reported as steers and split from the bulls into a 3rd contemporary group. If you commingle all three breeding groups, then the calves sired by bull A are out of cows that average roughly 1300 pounds and the calves sired by bull B are out of cows that average roughly 1400 pounds. If you assume that weaning weight proportional to 50% of mature weight, the result is that the calves sired by bull B should have an advantage of 25 pounds over calves sired by bull A (100 lb. difference in cows mature weight*50% adjustment for weaning weight)/2 because of one parents influence). Here is were it gets tricky, if you want to collect information about the cows then it would be best to have one group for each bull used. However, if you want to collect information about the bulls then only the cows that are randomly mated can be used for that comparison. To conclude, the practice of using small comparison groups could be an honest attempt to eliminate biases from entering TPR, not an attempt at chicanery. Frankly, if I were to cheat I would just fudge the numbers.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Oct 27, 2013 17:05:30 GMT -6
I visited a performance breeder some years ago. What he was doing was using the hot bulls with the top epds to compare to his home raised bulls. The calves by the hot bulls looked awfull. I don't know why they looked so bad, but it was obvious he was using them to build his home raised bulls' epds. I think the epds that don't hold up over time were caused by guys fudging the numbers or somehow stacking the deck to get the ratios they want against those proven bulls.
As for your example of the different breeding groups, if both parents have epds the formulas should weight the calves epds based on both parents epds. That is if you believe in the formulas.
|
|
|
Post by bookcliff on Oct 27, 2013 20:18:29 GMT -6
Tom, I really don't see how checking the milk volume on a few cows would allow them to estimate milk production across all blood lines in the Hereford breed. Bet it was an interesting project though, but with Kstate facilities that eliminated a lot of the danger! agreed. the one thing Jay told me is that in terms of what they saw at his dads, they were of the opinion butterfat content was more important than volume of milk when looking at it. and that also goes back to something my dad always said, back in the days when nurse cows were still legal on the show road ,a jersey or guernsey always made a better nurse cow than a holstien.
|
|
|
Post by lazyycross on Oct 29, 2013 9:15:54 GMT -6
|
|
|
Post by Mickelson on Oct 29, 2013 10:07:15 GMT -6
Now make sure those who attend the AHA annual meeting have page 13 of the BIF document memorized. I feel bad for Heather Bradford, she has to present on the new EPD. I know we have been through this before but...my opinion on this deal is as follows: someone is trying to address udder quality and that is great(much needed). The problem lies when breeders start manipulating the data being reported. Maybe we should go to a hand test. Back in college that's how I used to judge a lot of udders.
|
|
|
Post by bookcliff on Oct 29, 2013 13:44:09 GMT -6
Now make sure those who attend the AHA annual meeting have page 13 of the BIF document memorized. I feel bad for Heather Bradford, she has to present on the new EPD. I know we have been through this before but...my opinion on this deal is as follows: someone is trying to address udder quality and that is great(much needed). The problem lies when breeders start manipulating the data being reported. Maybe we should go to a hand test. Back in college that's how I used to judge a lot of udders. teat size or just suspension?
|
|
|
Post by herefordguy on Oct 29, 2013 15:55:39 GMT -6
|
|
|
Post by Mickelson on Oct 29, 2013 16:46:02 GMT -6
Now make sure those who attend the AHA annual meeting have page 13 of the BIF document memorized. I feel bad for Heather Bradford, she has to present on the new EPD. I know we have been through this before but...my opinion on this deal is as follows: someone is trying to address udder quality and that is great(much needed). The problem lies when breeders start manipulating the data being reported. Maybe we should go to a hand test. Back in college that's how I used to judge a lot of udders. teat size or just suspension? Suspension is key. Maybe there should be a "perky" category...I better stop there.
|
|
|
Post by picketwire on Oct 30, 2013 0:50:57 GMT -6
Started to try to add something to the conversation, but just couldn't do it justice now . . .
Will look forward to some thoughts after everyone gets the refresher course reviewed in frustrating mathematical formulas. Thanks for the link to see just exactly what the current (updated from when I last reviewed it) formula is and an even bigger thanks for reminding me why I did not have a future in mathematics! struggled thru statistics (instructor problem, not the work itself) but valued what I learned (about both statistics and idiotic political and bureaucratic secondary educational maneuvering).
|
|
|
Post by strojanherefords on Oct 30, 2013 15:53:56 GMT -6
Started to try to add something to the conversation, but just couldn't do it justice now . . . Will look forward to some thoughts after everyone gets there refresher course in frustrating mathematical formulas. Thanks for the link to see just exactly what the current (updated from when I last reviewed it) formula is and an even bigger thanks for reminding me why I did not have a future in mathematics! struggled thru statistics (instructor problem, not the work itself) but valued what I learned (about both statistics and idiot political and bureaucratic secondary educational maneuvering). I have the same thoughts and have had the same experience. EPDs and their accuracy figures should be expressed in a format that is easy to understand by the lay person. I think the best way to do this is by changing the way that EPD accuracies are expressed to a standard deviation approach: There is a two-thirds probability that the true EPD is between X units of the stated value provided that the information received by the breed organization is correct. Expected Progeny differences do not and cannot take into account variances in prepotency do to the genetics of the dam or environment.
|
|
|
Post by George on Oct 30, 2013 21:25:18 GMT -6
Jared: Have you identified any putative mutations in Herefords yet? Is it near the 76 you identified in Angus? How many Hereford sires whole genome sequenced in the research?
|
|
|
Post by herefordguy on Oct 30, 2013 21:51:40 GMT -6
Started to try to add something to the conversation, but just couldn't do it justice now . . . Will look forward to some thoughts after everyone gets there refresher course in frustrating mathematical formulas. Thanks for the link to see just exactly what the current (updated from when I last reviewed it) formula is and an even bigger thanks for reminding me why I did not have a future in mathematics! struggled thru statistics (instructor problem, not the work itself) but valued what I learned (about both statistics and idiot political and bureaucratic secondary educational maneuvering). I have the same thoughts and have had the same experience. EPDs and their accuracy figures should be expressed in a format that is easy to understand by the lay person. I think the best way to do this is by changing the way that EPD accuracies are expressed to a standard deviation approach: There is a two-thirds probability that the true EPD is between X units of the stated value provided that the information received by the breed organization is correct. Expected Progeny differences do not and cannot take into account variances in prepotency do to the genetics of the dam or environment. I too have thought that it would be easier to understand if EPDs were presented as confidence intervals.
|
|
|
Post by herefordguy on Oct 30, 2013 22:10:01 GMT -6
Jared: Have you identified any putative mutations in Herefords yet? Is it near the 76 you identified in Angus? How many Hereford sires whole genome sequenced in the research? We have finished the sequencing of 9 Hereford bulls. We are 1/4 of the way through the analysis of the sequence for 3 bulls. I previously misspoke. We identified 176 possible lethal loss-of-function mutations in the 11 Angus sires. We identified 1,720 likely loss-of-function alleles in 1,477 genes. Keep in mind that we are only in the start of the second year of a five year project. It will be quite a while till we have deliverables to the industry.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Oct 31, 2013 7:06:14 GMT -6
Anything is "possible".
|
|
|
Post by lazyycross on Oct 31, 2013 12:11:53 GMT -6
A confidence interval is a statistical measure for populations, not individuals.
The reason the "possible change" isn't printed as a "confidence interval" is because the calculation of EPD is not conducive to directly using confidence intervals. It is a statistical impossibility to have direct confidence intervals for EPD. Confidence intervals are designed for "hard" numbers with actual, simple measurements from populations, not a value that is derived as a expectation on a single animal, with ever-changing values for the EPD, prediction error variance, and breeding values for the entire population.
FWIW, the possible change values calculated by breed associations do use similar concepts to confidence intervals.
In effect, they are 68% (I believe this is generally correct) confidence intervals, based on the accuracy of the EPD.
Directly from the AAA: For example, a sire with an accuracy of .65 and birth weight EPD of +1.0 is expected to have his "true" progeny value falling within ±.92 pounds for birth weight EPD (ranging between +0.1 and +1.9) about two-thirds of the time. With the conservative approach taken with respect to heritabilities in the Angus evaluation, actual EPD changes of animals within the population are much less than statistics would indicate.
From the Hereford page: Changes in EPDs can be expected to fall within the possible change interval 68 percent of the time. In actuality, Hereford EPDs typically change considerably less than this level.
What you guys are asking for is already in place.
Dave K.
|
|
|
Post by George on Oct 31, 2013 13:23:43 GMT -6
From the Hereford page: Changes in EPDs can be expected to fall within the possible change interval 68 percent of the time. In actuality, Hereford EPDs typically change considerably less than this level. Dave K. Interesting statement! Care to provide the study or research that backs it up?
|
|
|
Post by lazyycross on Oct 31, 2013 15:52:26 GMT -6
No need, George.
Linear equations are used to produce the EPD; as such, they are relatively predictable. When the heritability of the trait is estimated conservatively as is generally the case in National Cattle Evaluation, one result is that the EPD will change less than predicted by the possible change.
This isn't magic; it is the direct result of how the EPD are formulated. It doesn't need to be studied to be verified; it is a consequence.
They worded it "softly"; the real case is actually more firm.
Dave K.
|
|
|
Post by George on Oct 31, 2013 17:10:02 GMT -6
So what you're saying is if the "possible change" parameters are set widely enough (ex. approaching Infinity) then there's at least a 68% chance the actual EPD will fall within the parameters.
No, I guess there wouldn't be any need to study that!
Also, no need to believe the EPDs in unproven bulls and cows with low accuracy are anywhere near the actual ACCURATE level of their true performance.
Just further proof of the shell game.that's being played.
|
|
|
Post by picketwire on Nov 1, 2013 9:54:55 GMT -6
If it helps, one might think of the possible change in this manner. Envision if you will the standard bell curve that helps illustrate statistical measures. The high point of the bell curve shows the average/mean of the population. Left side shows below average, right side shows above average. Statisticaly speaking half and half. A standard deviation from each side of that bell curve represents an additional percentage of that population. If I remember correctly two standard deviations together represent approx. 67-68% of said population. Same thing correlates to the range of expected movement within a parameter measuring whatever trait. The old sire summaries (printed) used to show the expected range of movement within a certain accuracy for each trait. The more accuracy a sire has the smaller the range of expected movement. The same thing works to shrink the range of each standard deviation measured within that said population depending on each individuals ability to transmit the gene or number of genes necessary to influence that trait in whatever direction you desire. The wider and flatter that bell curve is, then the more range you will see in actual measurements for that trait. A taller and narrower bell curve relays less range from top to bottom(right to left) for that trait being measured. The parameters are determined solely by the numbers submitted for analysis and the accuracy reflects that range over which all of those numbers fall, thus when more numbers are added and a sire has the ability to positively affect change in the direction desired, then that bell curve becomes taller and narrower as that grouping 'tightens' so to speak. On the flip side if you have an 'outlier' i.e. one who lies outside of normal plot of the curve then it does take longer and more numbers to tighten that bell curve up. Compare the bell curve of sire A with the bell curve of sire B and you begin to see the differences between the two. Now the BLUP formula that figures the EPD takes these differences and numbers and comparisons of each individual, parents, sibs and contemporaries and heritabilites and kicks out a number that far too many people take as gospel without fully understanding either the formula (which is very difficult to understand without a solid math background) or the population dynamics (i.e. contemporary grouping) making up the population. The process is the best we have right now and I am certain that smarter people than me mathematically can explain the rest of it better than I ever will, but it does have me grounded in the understanding of how it works and why without proper grouping and 'honest' reporting of data, be it either intentional or accidental, it will not work any better. The entire process is steeped in statistical analysis that doesn't allow numbers to be fudged or parameters set widely enough, but the data going in can be garbage data that will not allow an accurate picture to become possible.
I do wish there was a way I could draw the bell curve on this post to help illustrate what it is I am talking about because it really does help 'see' the plot of the data and the standard deviations.
|
|
|
Post by Glenn on Nov 1, 2013 10:07:52 GMT -6
|
|
|
Post by lazyycross on Nov 1, 2013 10:17:16 GMT -6
George,
You misunderstand. The way you worded this, it is as if the possible change values are set and then we have a particular level of certainty about whether the true value is contained there. This is precisely opposite of what is happening. We choose the level of certainty 1st, and the possible change parameters are the result of that choice. What you have suggested, although backwards is that if we choose a 100% certainty, the possible change parameters would be very wide. That is correct. However, it is an exercise in futility, because it should be obvious that we would be 100% (not 68%) certain that any particular value actually does occur in the population it came from. In general, we would chose a level of 99% certainty, rather than 100%.
This is relatively complex algebra, George. Most people are not exposed to this level of algebra as undergraduates unless they are statistics majors. Universities that specialize in Animal Breeding PhD's are beginning to teach this level of algebra to Master's students, but most of the understanding doesn't really come until the PhD level. I've only met a handful of seedstock producers in the last 20 years that can converse at an articulable level regarding this level of algebra, maybe 4, actually, but really only 2, that REALLY could. That isn't a statement of disrespect, and in no way disparages the knowledge of producers, it is just a simple fact that we are talking about "high" math, and it takes specific training to understand.
I trust that A*B = AB and that 0.5A*B = 0.5AB, and that 0.5AB will ALWAYS be less than AB. The formula for accuracy that is partially derived from heritability, that ends up in the possible change equation is the same concept, only with higher level math. An overestimated heritability will lead to a decreased possible change range and greater probability of the true BV being different than the EPD. The other consequence of an overestimated heritability, is that the EPD spread will be wider, which can be erroneous as well; this would overemphasize small differences among animal leading to even more "bigger is better" mentality.
Dave K.
|
|
|
Post by lazyycross on Nov 1, 2013 10:27:34 GMT -6
picketwire,
The range of observations on an animals progeny have nothing to do with how fast the accuracy of the EPD increases, nor does it have anything to do with the possible change value for an animal.
Yours is a common misunderstanding.
The taller, narrower bell curve you described from the older sire summaries was a graphical representation of how the possible change gets narrower with more observations, not to be confused with a narrower range of performance in the progeny of the animal.
To take an extreme example. If bull A's 3 progeny weighed 400, 500, and 600 pounds at weaning, and bull B's 3 progeny weight 475, 500, and 525 pounds at weaning, the sire's accuracies and possible change values would change at the same rate given everything else being equal.
Dave K.
|
|
|
Post by Glenn on Nov 1, 2013 10:28:15 GMT -6
EPD bulls
Work good for these guys. Probably what I need to be doing. Probably sell bulls for 4x or better than what I do. Sell bulls to the studs...ect...
My dumb ass can't get past how they walk though....
|
|
|
Post by Glenn on Nov 1, 2013 10:37:36 GMT -6
Here's a term that I learned in 7th grade back in 1982. No high level degrees required. First day of computer programing. (yes back in the day they taught 7th graders BASIC)
GIGO
|
|