RE: evolution and diet (was: FITNESS: Diet and Exercise)

From: gts (gts_2000@yahoo.com)
Date: Mon Apr 21 2003 - 09:06:57 MDT

  • Next message: Dehede011@aol.com: "Re: LUDDISM: "Good Life" debunked"

    Eliezer S. Yudkowsky:

    > Unfortunately, the statistics you have learned is the false way.

    Oh really?

    > Least squares and p-tests are but a lie and a deceiver. To learn the
    > true way you must unlearn the old knowledge and study Bayes.

    Just how old do you think I am, Eliezer? I studied Bayes also. I assure you
    I was not educated prior to the 18 century! :)

    The principles I mentioned in my posts were not about any particular
    statistical test (there are dozens of them) or about bayesian vs
    frequentists interpretations of statistics. I was just espousing some very
    basic principles of scientific research as exemplified in statistical
    health-related research but also obvious is all scientific research.

    With respect to "burden-of-proof table tennis", the main bone of contention
    here, the burden of proof is always on those who would challenge the status
    quo, i.e., it is always on those who would offer some new theory that they
    think should better explain the empirical data.

    A classic case is Galileo and his attempt to prove a Copernican heliocentric
    model of the solar system against the then widely accepted geocentric model.
    In basic statistical research terms, the geocentric model was the "null
    hypothesis," or "working hypothesis," accepted as valid and true by
    virtually everyone at the time, while the heliocentric model was the
    "competing hypothesis." The burden-of-proof was on Galileo to prove the
    competing hypothesis -- it was not on the scientists of his day who
    believed, along with the church, that the Earth was at the center of the
    solar system.

    If you don't think Galileo had the burden of proof then you haven't read the
    history of Galileo, and about the great burdens placed on him in his effort
    to prove the heliocentric model and disprove the geocentric model.
    Fortunately we are not so dogmatic or hysterical about science in this
    post-enlightenment era, but the burden-of-proof concept is still a reality
    of modern science, as it should be.

    A more modern example: the burden-of-proof was on Einstein to show that
    Newtonian physics is incomplete. Newton was under no obligation to rise from
    the grave to show that Einstein was wrong. Newton was the incumbent.
    Einstein was the new contender who needed to prove his theories against
    Newton's.

    As I wrote in another post, the most successful and reputable scientists,
    like Einstein and Galileo, are those who gladly shoulder the
    burden-of-proof. A scientist gets an idea that the current widely accepted
    working hypothesis is not as accurate a theory of reality as some other
    theory, and decides to expend a lot of his time and energy and resources to
    design and conduct an unbiased experiment to test whether that the other
    theory explains the data better than the old working hypothesis. Those who
    do so successfully advance scientific knowledge for all the world, and in so
    doing they earn the respect and admiration of their peers -- a key
    motivating factor for people in academia. This is true regardless of the
    veracity of Bayesian theory.

    With respect to this thread, my point is simply that according to me and
    other people who embrace paleodiet theory, paleodiet theory really ought to
    be the null (working) hypothesis for optimal diet. From your posts on the
    subject in which you state words something like "I tend to think ancestral
    diets should be the working assumption until proven wrong" I think you agree
    with me, Eliezer. You are saying the same thing I am saying, but in less
    formal language.

    As an aside, sometimes in stat we don't use the term "null hypothesis" in a
    literal way, for example when testing two competing theories in a field in
    which there are no pre-existing theories, or when comparing two samples of
    data expected to have different variances, but in greater terms the
    principle is still there: the null hypothesis in such cases is that neither
    theory explains the data better than the other.

    -gts



    This archive was generated by hypermail 2.1.5 : Mon Apr 21 2003 - 09:15:35 MDT