Quote:
Originally Posted by Rise Over Run
What is the statistical significance of 100? Other than being an easily identifiable number, I'm guessing none.
|
There is no statistical significance to 100.
The higher the number of subjects for a study the more reliable the data is from parametric statistics.
Having said that there are instances when only very small numbers are available and non-parametric statistics have to be relied on.
For example, in cases of very rare diseases, a clinician may only be able to study a handful of people in the world.
The belief is that the use of larger cell sizes, the chance of random factors intervening lessens. Arguments can be made against that idea to.
It is the nature of parametric statistical tests that the larger your sample size, the easier it is to reach a .05 level of significance or lower.
That looks good in research papers, but it is partly an artifact of the confidence level of the statistic itself.
In the case of horse trainers, many do not have steeds that race 100 times a year.
That doesn't give them an escape route, if you buy the idea that I do, cell sizes can be as low as 25 and you can see patterns forming that will likely be significant if it continued through a 100 or many more trials.
Having said that, statistics, while interesting cannot convict this trainer.
In an earlier thread I said that the change of her horses winning so frequently before the vet was caught, and after the vet was caught were .049 on a Chi Square test.
That's an interesting fact and does throw a cloud of suspicion.
But really, it does not demonstrate her guilt.
Let's face it.
The vet was caught red-handed. That is a far better proof of his guilt.
It does not demonstrate that she was party to it, unless he starts "singing."