I remember doing a quick look one time and concluded a few thing. A
significant number of the metabolic pulls were on horses that were not
registered with the AERC. In that case all the pulls have the same ID -
none. Those would have to be thrown out but those are most likely the
horses with very few miles.
The one think I was looking was multiple metabolic pulls in a season -
that is how many horses had more than one metabolic pull in a season.
There were several with two - a fewer number with 3 and one or two with
4. Turns out they were mostly Arabs or half Arabs and one or two
weren't listed. Was that a surprise ( or significant ) neither since at
random given the percentage of Arabs and half Arabs the projected
number was not out of line with what was seen.
We need to start keep better records if we want to address these
issues. From what I remember in 2002 of the 20000 and change starts
these were about 4300 horses. When you go to sort the starts to the
4300 horses you find the horse with the most starts was "no ID" and
that was about 500. I just wonder if these 3 low mileage horses that
died were in this "no ID" category.
If you have some idea on how to best handle this and not skew the
conclusion, contact me privately.
But if you look farther the disproportionate number of
non-Arabians were also the horses with the least experience.
Got to be careful not to fall into the "a frog with no hind
legs can't hear" syndrome.
The total number of miles of these three horses was 125 miles.
I think you'd get a lot better numbers by looking at metabolic pulls. Take
all the metabolic pulls, put in horse age, experience, sex of horse, and
breed. Then compare that distribution against the overall population. Apply
a t-test to see if the distribution is significantly different, and then you
have some real data. This of course assumes that metabolic pulls and death
are related phenomena and stem from the same causes. Some metabolic pulls
(e.g., tying up) are probably not well correlated, so that will cause some
skew in the results. Reporting error also introduces error as well. I think
the much larger sample size is worth making these assumptions.
If I could get my hands on a few year's worth of data, I could do the
statistics. I'm still for turning this into a real study.
11 horses isn't a large enough sample to tell anything from.
