I’m kind of with Viney here, do these stamm numbers trace back to one mare or what?
At least with TB mare families you know what you are dealing with when somewhen says 9-c or 14-c or any of the founding mares from numbered family’s.
So if you say family 763 from the Hanoverian line it would be nice to say oh yeah that’s the founding mare so and so with a family that’s produced such and such.
Anyway, me chiming in.
Terri
Tri, I don’t know how other US studbooks regard the WBFSH, but the AHHA failed to send anyone to the very important meeting held in France earlier this month. Several BOD members were in Europe at the time and could easily have gone. Much information is lost to the membership when our studbooks take so little interest. But maybe other US based studbooks sent representitives… I don’t know. I was very disappointed to learn of this faux pas on the part of the AHHA.
These things are important for worldwide recognition of our horses. Campesino’s son was ranked 25th in the world on the showjumping list, making Campesino very well known throughout Europe; in the US… no recognition. I think we breeders have to stay on top of what is going on between our breed associations and the WBFSH or all may slip through the cracks. If our breed association doesn’t understand the importance of the WBFSH to US breeders, this is a huge problem for us.
Showjumpers, I agree wholeheartedly. It is once again, a problem stemming from the fragmented nature of our system here - a bunch of small, mostly poorly funded groups that backbite and bicker among themselves with no strong, cohesive direction, goals, or inititatives.
Do any of the wb registries even have a 1 year plan or a 5 year plan? Do you know what it is if they do? Is there any accountability? Is there any acknowledgment of goals attained? Is there any tracking of market percentage? Number of horses sold, average prices, etc? Is that info then marketed?
Very interesting info altogether. From all I gather the WBFSH-ranking works similar as the German FN rankings used to work before the BLUP was fashionable. The biggest point of critizism back then was that the recognized (rated S-level shows) show result had too much of an environmental factor to them and that anything that was not rated S-level classes fell under the table making for often misleading statistics.
Today we see the opposite extreme within the FN-breed value calculation as many young stallions score major brownie points for their own merits in young horse performanc tests (e.g. BC-titles, stallion performance test results) which are as we all know fairly political by nature (no objective measure such as time and penalties but humans judgeing for the most of it) whereas some longstanding proven producers are downright kicked out of the statistics no matter how many top class performers they have produced over the years (Raphael comes to mind who’s showjumping inheritance is certainly not argued a lot amongst showjumper folk?)
While I clearly agree top level competition should play an important role in breed-value calculation or any breed ranking for that part, I don’t think it is wise to fade out everything that happens in the sport at width because not only is the vast majority of riders competing in the lower levels but also this is where the biggest marketvalue lies. It is not in the top10 let’s face it. Most breeders will never make one of those six figure or up sales so I think we’d be fools to base our breeding decisions on data that includes topsport only. To me the most valuable line is the one I can be sure will not produce ‘complete failures’ because it’s those complete failures that cost a breeder money. Everything else can still be turned for the better and if I end up with one that can at best be a good lower level amateur mount he can still turn into a walking advertisement for my programme. Therefor I look for more broad-based data to found my breeding-ideas upon. Still very good information to have at hand.
As far as plans go I can say within the Hanoverian Verband there are quite a few goals definied. The guys at Verden work very hard yearround and the statistics of how many they marketed for how much and where are very easily accessed. There is a huge annual report published for everyone to follow and after each major breeding event the board members will comment at length and get guest commentatory as well. While there is nothing like a public 5year plan one can certainly say there are goals defined and followed and I can say quite successfully achieved.
Interesting logic. If the rankings were meaningless, you say, they would not be compiled. So the fact that they are compiled means they must be meaningful.
Hmm.
Note that Decartes wrote “I think therefore I am.” He did not write “I am therefore I think.”
This list could be more informative, but it is not meaningless. Reece agreed with you that it could be more telling by using the per 100 method. Do I detect a hint of resentment here?
I wonder if you would have the same opposition if the Irish Sporthorse or KWPN had the number one ranking.
Oh Kareen, I know that they have it in Europe. I should have been more clear in that I was referring to the wb registries operating in the U.S. and as far as the North American marketplace goes.
The european registries as it pertains to EUROPE have a very organized, strong business model that has proven successful for many years.
Of course I would have the same critique. I have similar critiques about the WBFSH rankings in which the KWPN is first in jumping and the IHB is first in eventing. In fact I wrote an article in Horse International magazine (also on my blog) that discusses the uses and misuses of index scores and offers several suggestions for improving the Irish and Dutch indices and the WBFSH rankings.
In an statistical analysis if the data are bad the analysis is worthless.
In any statistical analysis if the statistical model is wrong (or its assumptions violated) the analysis is worthless.
Some of us on this board have advanced degrees in disciplines that require advanced statistics and research methodology. The errors and misjudgements in this ranking and in the studbook indicies and WBFSH rankings are so basic that one does not need advanced training to shoot holes in them a mile wide.
Google “sampling on the dependent variable” and in five or ten minutes you will learn one important fatal flaw in this type of ranking.
These rankings are compiled by the Dutch and the Germans. They obviously have some belief in them. Maybe while you’re telling these people how wrong they are… and how right you are… you will go down in history as another Tessio. I’m sure they appreciate your good intentions.
So the authors believe in their rankings. Big deal! It does not make their analysis correct or useful.
I note that you have not made one statement arguing WHY you believe these rankings are correct and useful from a statistical and/or methodological point of view.
Signing off…
I love this discussion so far! (okay, not the infighting but…)
While I can’t follow all the nuances of how the data presented here is collected/tabulated, as an epidemiologist, I agree with whoever said that any “result” is only as good as the data put in: if the data is crud, then so will be the results. See it in medicine all the time. Anyone can design a study to prove a desired result: a careful study, with a true null hypothesis, and great design, can withstand any assault.
I think I read that there may not be enough controls on the variables affecting the data: what are the skew factors? ( I think Kareen hinted at some of them).
I have no clue :winkgrin:, I just know I spend a lot of time looking at data, and “p” factors :lol:
The most fundamental problem with the analysis is that they are sampling on the dependent variable – international showjumpers. They have a list of international showjumpers (their dependent variable) and they are sorting the showjumpers into their respective motherlines. Then they are ranking the motherlines by the number of international showjumpers each motherline has produced.
What is wrong with this approach? You are an epidemiologist so let’s use an example from your field. Let’s say there has been a large increase in brain cancer deaths in a jurisdiction and you are asked to investigate it.
An inexperienced researcher will develop a study that contains, let’s say, all the people who died of brain cancer and then try to identify causes. The inexperienced researcher notes that 98.8% of the people who died of brain cancer watched Monday Night Football and this variable is the one that the greatest number of brain cancer patients shared. There you go! Monday Night Football must have caused the brain cancer!
But of course that is a silly conclusion.
A more appropriate technique woud have been to have drawn a random sample (of some type, a purely random sample would not provide enough cases of people with brain cancer since the incidence of brain cancer is still low in the population despite the spike in cases) to estimate a model that attempts to discriminate (to sort) between those with brain cancer and those without. And you could use a statistical estimator such as discriminant analysis or logistic regression (died of brain cancer = 1 and not developed brain cancer = 0) to test your theory. And if your theory is good, and your data are good, and you have enough statistical power you might find one or more variables that are important predictors/explainers, from both a substantive (remember Monday Night Football?) and a statistical point of view, of why certain people died of brain cancer and others did not. And included in that model would have to be control variables such as age, sex, race, etc. – variables that may be correlated to your key independent viables but are themselves certainly not possible causes of brain cancer.
In the case of these mare families we have a similar problem of sampling on the dependent varable. The authors took as their population to be studied the population of international showjumpers (i.e., the brain cancer casualties). They should have taken as their population to be studied the entire population of each mare family (i.e., the brain cancer casualties AND the people without brain cancer) and then sorted each mare family into those horses that are/were international showjumpers and those that are/were not. So the solution is so much easier here with the mare families than it is with the cancer study: all they have to do is control for the size of each mare family! Simple.
Let me give another example. Ireland (was and maybe still is) the second largest exporter of software in the world after the USA. So the USA is like the most successful motherline in the study and Ireland is like the second most successful motherline in the study. But wait a minute. Ireland has only about 1.5 percent of the total US population. So on a per capita basis Ireland is a much more “powerful” exporter of software than the USA.
Or consider whose record is more impressive in producing Nobel Laureates in Literature? Ireland has one-third (4) the number of Nobel prizes in literature as does the USA (12). Who impresses you the most by their production of Nobels in Literature? Ireland or the USA? For me the answer is easy: Ireland.
Let’s go back to the mare families. For this ranking to have any utility we have to know the denominator, how many horses are in each of the mare families. We know stamm 776 is big. But how big is it? If we took into account the size of each of the mare families would the ranking be the same? Holding size of the mareline constant does the ranking stay the same? Or do the positions shift dramatically once we take into account how many horses are in each mare family?
Is stamm 776’s production of international showjumpers impressive? I do not know and the only way to truly know is if we control for size. And the most reader-friendly way to do that is to compute a rate – number of international showjumpers per 100 horses in each motherline.
Without controlling for the size of the mare family and/or computing a rate we simply have no idea which mare families are truly powerful producers of international showjumpers and which ones only appear to be so because they are so big.
Tom - I am of Irish descent. Don’t need to impress me with Ireland. I just wish they hadn’t become the largest exporter of software, then the Shelbourne might be the wonderful old hotel it was, not the modernized hotel it is now (owned by Marriot!). PS: my grandfather brought the Abby theater to America…
From what you have posted, the analysis is backwards: it appears to have started with the “end” product and worked backwards. However, I don’t see how not to start with the end product - how else to identify a place to start?
However, once identified, ie top marelines, you suggest, and I buy it, that the next step would be: of all offspring, of said mareline (identified as having produced “a” or “several” GP jumpers), how many offspring of said mareline, with number of offspring the denominator, have produced GP jumpers? 1 in 10? 1 in 50? 1 in 100? And can you control for number of shows?
Or, perhaps more critical, as Kareen points out, what about the offspring who excel below the radar? Is that not perhaps even more important than the 1 in 100 GP jumpers???
I am sure I am regurgitating what you have all ready posted Tom, just have to work it through my tiny brain cells.
In any event, love the TV and brain cancer analogy. We have had many a correlation “proved” - equally absurd - only to have to retract years later. Luckily, everyone is still watching football :lol: go Redskins!
So these ranking are from just this last year – are they tail-female lines?
I think to be valid, like with stallions, you have to look at percentages.
If stallion A has produced 10 FEI horses, and Stallion B has produced 2, who is the better sire?
It seems like Stallion A, but if Stallion A has 2000 offspring of riding age, and Stallion B only has 100, then Stallion A has .5%, and Stallion B has 2%, then I would go with Stallion B. To take that a step further though, you would have to know how successfull of producers the mares were that had been bred to each, ages of the offspring, and how many offspring found themselves in a trainer’s barn that could produce an FEI candidate. and then there is the financial aspect, the match of horse to rider, etc, etc.
There is just no easy way to make a valid and meaningful list, but they are lots of fun to use for bragging. 
What I DO know is that if you breed Luna to Salieri you are likely to get one heck of an athlete! :yes:
(Even more interesting to me as a breeder of older type blood is that there is no real refining blood for many generations, but the offspring are modern type) <I really just love the Sender, Don Carlos, Lungau blood)
[QUOTE=tom;3692513]
The most fundamental problem with the analysis is that they are sampling on the dependent variable – international showjumpers. They have a list of international showjumpers (their dependent variable) and they are sorting the showjumpers into their respective motherlines. Then they are ranking the motherlines by the number of international showjumpers each motherline has produced.
What is wrong with this approach? You are an epidemiologist so let’s use an example from your field. Let’s say there has been a large increase in brain cancer deaths in a jurisdiction and you are asked to investigate it.
An inexperienced researcher will develop a study that contains, let’s say, all the people who died of brain cancer and then try to identify causes. The inexperienced researcher notes that 98.8% of the people who died of brain cancer watched Monday Night Football and this variable is the one that the greatest number of brain cancer patients shared. There you go! Monday Night Football must have caused the brain cancer!
But of course that is a silly conclusion.
A more appropriate technique woud have been to have drawn a random sample (of some type, a purely random sample would not provide enough cases of people with brain cancer since the incidence of brain cancer is still low in the population despite the spike in cases) to estimate a model that attempts to discriminate (to sort) between those with brain cancer and those without. And you could use a statistical estimator such as discriminant analysis or logistic regression (died of brain cancer = 1 and not developed brain cancer = 0) to test your theory. And if your theory is good, and your data are good, and you have enough statistical power you might find one or more variables that are important predictors/explainers, from both a substantive (remember Monday Night Football?) and a statistical point of view, of why certain people died of brain cancer and others did not. And included in that model would have to be control variables such as age, sex, race, etc. – variables that may be correlated to your key independent viables but are themselves certainly not possible causes of brain cancer.
In the case of these mare families we have a similar problem of sampling on the dependent varable. The authors took as their population to be studied the population of international showjumpers (i.e., the brain cancer casualties). They should have taken as their population to be studied the entire population of each mare family (i.e., the brain cancer casualties AND the people without brain cancer) and then sorted each mare family into those horses that are/were international showjumpers and those that are/were not. So the solution is so much easier here with the mare families than it is with the cancer study: all they have to do is control for the size of each mare family! Simple.
Let me give another example. Ireland (was and maybe still is) the second largest exporter of software in the world after the USA. So the USA is like the most successful motherline in the study and Ireland is like the second most successful motherline in the study. But wait a minute. Ireland has only about 1.5 percent of the total US population. So on a per capita basis Ireland is a much more “powerful” exporter of software than the USA.
Or consider whose record is more impressive in producing Nobel Laureates in Literature? Ireland has one-third (4) the number of Nobel prizes in literature as does the USA (12). Who impresses you the most by their production of Nobels in Literature? Ireland or the USA? For me the answer is easy: Ireland.
Let’s go back to the mare families. For this ranking to have any utility we have to know the denominator, how many horses are in each of the mare families. We know stamm 776 is big. But how big is it? If we took into account the size of each of the mare families would the ranking be the same? Holding size of the mareline constant does the ranking stay the same? Or do the positions shift dramatically once we take into account how many horses are in each mare family?
Is stamm 776’s production of international showjumpers impressive? I do not know and the only way to truly know is if we control for size. And the most reader-friendly way to do that is to compute a rate – number of international showjumpers per 100 horses in each motherline.
Without controlling for the size of the mare family and/or computing a rate we simply have no idea which mare families are truly powerful producers of international showjumpers and which ones only appear to be so because they are so big.[/QUOTE]Been reading this forum off and on for what, 10 years… or however long it’s been around… this discussion is one of the most useful ever. Thank you.
Another problem, as Tom’s analogy makes clear, is that new stamm numbers with much less chance to produce (time being what it is) are still being created. That would have to be another skew factor when listing families by stamm number.
At least in TBs, the root families go back to the seventeenth and eighteenth centuries, and all of the lettered subfamilies eventually go back to one of the root mares, leaving MtDNA analysis out of it.
Or consider whose record is more impressive in producing Nobel Laureates in Literature? Ireland has one-third (4) the number of Nobel prizes in literature as does the USA (12). Who impresses you the most by their production of Nobels in Literature? Ireland or the USA? For me the answer is easy: Ireland.
I like your analogies with the exception of this one. The awarding of Nobel prizes is subjective and very political. Keep in mind that they gave that idiot Al Gore a Nobel prize. Also, the committee that awards these prizes know that they can’t load up the U.S. with prize after prize - there is already enough hostility towards us around the world.
Your points also remind me of the threads regarding the Hilltop stallion, Riverman. He has bred 100s of mares - rumored at 200 mares a year? Yet, few have done much of anything. Pro Riverman posters point to a horse here and there, but from all those breedings, you would think there would be lots at all levels of sport. There isn’t.
By submitting your email, you agree to our Terms of Service and Privacy Policy. You may also receive promotional emails from The Chronicle of the Horse. You can opt out at any time.
For Customers
Company
Terms of Use
