>From: [EMAIL PROTECTED]
>Subject: Re: t-and-f: Performances comparison
>
>In a message dated 11/7/00 11:25:34, [EMAIL PROTECTED] writes:
>
><< This website, developed by Rinaldo Zocca, has developed a
>statistically-sound approach to comparing world records. MJ's 400 and 200
>are ranked as the most advanced, even compared to FloJo's 100, which is the
>leading women's performance.
>
>http://digilander.iol.it/rzocca/reccomp/reccomp.htm >>
>
>Sorry, but i have trouble accepting "statistically sound" for any methodology
>which rates anything above the 10.49. Not only was the legal-wind flow on a
>different planet from the rest, but the 10.49 was also not legal, pushing it
>even farther out of the realm of probability (how windy was it? let's just
>say that as the race ended i looked up and saw Dorothy and Toto go whizzing
>by).
>
>gh
The comparisons between men and women may be skewed by a simple fact: the
women's fields are less "dense", i.e., that competition is simply not as
close as for men. I've made the point earlier on this list that women
elites simply don't have the same number of competitors at the same level
as men. I think I made a comparison of performances in a couple yearly
performance lists showing this. If the performances are more spread out,
then the standard deviation increases. It doesn't matter which direction
the performance spread increases. Thus, the fact that the women's lists
must go to relatively slower times at the 500th position means that
extraordinary performances are "closer" to the mean as described by the
variance statistic, a somewhat counterintuitive result.
The first fix might be to describe the distribution using a log-normal or
some other distribution, rather than the normal which is probably the case
here. Comparison to the third moment, skewness, of the distribution might
also be useful.
Having said this, the statistical validity within genders is still
strong. The underlying problem may be that the women's performance list is
somewhat restricted by the fact that they have only recently achieved
roughly the same competitive opportunities as men. Another problem is that
the analysis is looking at only the tail of a much larger population
distribution for which we have virtually no information. In the end, we
may not be able to make comparisons between the genders because the
population characteristics are not similar enough.
I found that MJs 400 surpassed the 200 somewhat of a surprise. Note that
the men's 10k is only about 5th on the list.
Richard McCann
