Re: Hope for Pitman's Statistics

brogers@xxxxxxxxxxxxxxxxxx wrote:
>Well, not really. Sean was keen to show that a guy with a simple
>statistical idea could overturn the entire field of evolutionary
>biology. Not very likely, especially with a very bad statistical idea.
>A recent article in Science does, however, basically set a whole little
>field on its heels with a simple statistical idea that had been
>overlooked (Nee et al [2005] The illusion of invariant quantities in
>life histories. Science 309:1236-1239).

>A mini-field had grown up over the past 10 years trying to understand
>"invariant quantities" in life histories. These invariant quantites
>were things like the ratio of age at sexual maturity to average
>lifespan or of weaning weight to maternal weight. These dimensionless
>ratios were allegedly conserved across many taxa and over many orders
>of magnitude and the correlation coefficients and R squared values
>between, say, weight at weaning and maternal weight, were remarkably
>good for biology, up in the 0.90-0.97 range. A bunch of papers were
>published trying to figure out the deep unifying principles of biology
>that accounted for these dimensionless constants in life histories.

>The authors of the new Science paper, however, noticed a simple
>statistical point. All the ratios are constraned to the range 0-1,
>while the original values (age or weight or what have you) vary over
>several orders of magnitude. It was pretty easy to show that under
>those circumstances you inevitably get a very high R squared value,
>even if the ratios are randomly distributed between 0 and 1. It's
>pretty clear that the whole "invariance" in life histories was a simple
>statistical artifact.

>Not that this is much comfort to Sean, but it does show that a bunch of
>smart people can be led down the primrose path because they overlook a
>relatively simple issue in statistics.

All true.

But the problem with Sean wasn't so much his statistics
as his model. He was assuming that to get a new
protein you had to take an old one and change it
residue by residue. And all the intermediate forms
had to be viable.

Of course it doesn't need to happen that way, but
you couldn't tell Sean that.

But he was right about his model. It really can't happen
that way.

Guesss what? It doesn't happen that way. If he'd been
just a bit smarter he'd have realized that he'd just
shown that.

----- Paul J. Gans

.

Relevant Pages

  • Re: Hope for Pitmans Statistics
    ... Sean was keen to show that a guy with a simple ... > lifespan or of weaning weight to maternal weight. ... All the ratios are constraned to the range 0-1, ... > relatively simple issue in statistics. ...
    (talk.origins)
  • Re: Hope for Pitmans Statistics
    ... Sean was keen to show that a guy with a simple ... >>> A recent article in Science does, however, basically set a whole little ... >>> lifespan or of weaning weight to maternal weight. ... >>> relatively simple issue in statistics. ...
    (talk.origins)
  • Hope for Pitmans Statistics
    ... lifespan or of weaning weight to maternal weight. ... good for biology, ... All the ratios are constraned to the range 0-1, ... relatively simple issue in statistics. ...
    (talk.origins)
  • Re: Statistical Pattern Recognition
    ... I didn't snip your wiki link ... as others have pointed out by doing exactly what Sean said they ... Sean said that he posted the Wiki ... It does not contain a Wiki link to statistics, ...
    (talk.origins)
  • Re: Hope for Pitmans Statistics
    ... Sean was keen to show that a guy with a simple ... >> lifespan or of weaning weight to maternal weight. ... All the ratios are constraned to the range 0-1, ... >> relatively simple issue in statistics. ...
    (talk.origins)
Loading