MINESHAFT

[TGJB]

Ergo the request for suggestions. Aside from the usual suspects, we know a very bright young guy with a degree in statistics and a serious interest in pedigrees who I\'m going to talk to about this-- as it happens  his father was also the guy who wrote the Bloodhorse article that got milkshakes banned in Kentucky.

As Patrick Cunningham said in the atricle I quoted, about 6% of colts are used for breeding, versus about 53% of fillies. What this means in practice is that a mid level stallion is really at the 97th percentile of available males, where a midlevel filly is at the 77th percentile of females-- there is much less range among stallions than the mares they are bred to, with the biggest difference between a 25k stallion and a 2.5k stallion being the book of mares they cover. So we are going to try to find a way to deal with the mares somehow-- anything based purely on sires will be misleading.

[Michael D.]

the 53% stat makes the project very difficult. we know that 53% of female horses do not get to the races, so you will have a very small sample of female figs to go on, compared with the number of mares. i will go back and read my old Vuillier and Varola stuff (hopefully today), and see if i can think of some way to improve on their methodology by using actual running figures.

I\'m not sure I have a complete understanding of the goal. If I were going to try to use the female side for statistics, I would \"start\" by categorizing only those fillies for whom I have past performances and perhaps speed figures.

The horses used in the following example are simply my attempt to make a point. They are also going to date me. :-)

I would put horses like Dainty Dotsie, What A Summer, My Juliet, and Safey Kept in one category for example. These are more or less \"highest quality pure sprinters\". If I had a large sample of so called \"highest quality pure sprinters\", I might then want to break them down according to their own sire\'s characteristics and form sub-categories within that group.

I could then for example look at all the offspring of the \"highest quality pure sprinters\" sired by \"pure sprinters\" and form conclusions based on that sample.

That would certainly be more useful than looking at 2-3 offspring from a single filly.

The more stats I had, the more subcategories I could create. I might even consider incorporating multigenerational filly stats.

I\'m not sure how I would assign the proper weights between the sires and these grouped filly stats because the sire line stats would still probably be superior because they are direct. You may have to back into something like that with some type of  regression analysis.

I have no formal education in statistics and am thinking out loud as I type this, but I can\'t see how you can get around grouping the fillies in some way if you want results you can analyze.

I look forward to the solution. I hope my 2 cents is worth at least that much. :-)



Post Edited (11-21-03 16:21)

[TGJB]

Grouping by \"type\" doesn\'t address the issue of the genes being passed on in a particular pedigree, but that could certainly be a seperate study. I also don\'t know to what degree the figures of the dam herself will be of use, we will have to look at that. But it\'s been obvious when we looked at this that it is not a simple question, which is why we haven\'t done it already. Now that we have time we\'ll have a go at it, after talking to a number of people. Derby 1592, any thoughts?

By the way, we will obviously be looking at breeding \"nicks\" in conjunction with this.

>Grouping by \"type\" doesn\'t address the issue of the genes being passed on in a particular pedigree, but that could certainly be a seperate study.<

The groupings would at least have the  potential to be similar genes. It would make the sample sizes larger - especially if they could be refined beyond racing characteristics down to the fillies own pedigree (sire).



Post Edited (11-24-03 00:05)