From:    James J. Lippard
Subject: Re: sTARBABY to death

In article <1992Feb17.190422.5695@rosevax.rosemount.com>,
grante@aquarius.rosemount.com (Grant Edwards) writes...
>David.Rice@ofa123.fidonet.org (David Rice) writes:
>: Tag: Grant Edwards
>:
>: >GE> "I don't consider Gauqelin's (sp?) work double-blind, since he
>: >GE> seemed to arbirarily throw out data after examining it."
>:
>: What, specifically, leads you to believe (or understand) that he
>: had biased his data? The Gauquelin's published extensively their
>: data and methodology. Anyone may duplicate their experiments and
>: atempt to falsify the Gauquelin's results.
>
>According to accounts I've read, G's selection of "championship
>atheletes" was pretty hard to justify.  He included groups like
>private pilots, and threw out pro basketball players.  As I recall he
>also chose not to include people from certain geographical areas (was
>it Paris?).  He also refused to consider people born after a certain
>year (1950?).

The effect is only supposed to work with natural births, not drug-induced.
That's why he only collected births before a certain year.  Prior to the
CSICOP test of American athletes, Gauquelin said that he hadn't found the
effect for basketball players, and so they shouldn't be used in the test.
They were used anyway.  I don't believe Gauquelin ever placed any restrictions
on geographical area--he has plenty of data from Paris.  In Kurtz, Zelen, and
Abell's analysis of Gauquelin's data, however, *they* divided it up into
geographic subsamples and argued that the effect was present in Paris but
not in other regions.  (This division of data, however, was itself unsound.)

>The selection criteria seemed pretty contrived to me.

Some of it seems rather post hoc to me, too.  The data collected by
the French skeptics (CFEPP) should resolve these problems.  It should
also be noted that people have looked into the possibility that
data selection is responsible for Gauquelin's results and found some
minor problems, but nothing that accounts for the effect as a whole.
Suitbert Ertel's contribution to the proceedings of the Dutch European
Skeptics conference addresses this issue.

>--
>Grant Edwards                                 |Yow!  Go on, EMOTE!  I was
>Rosemount Inc.                                |RAISED on thought balloons!!
>                                              |
>grante@aquarius.rosemount.com                 |

Jim Lippard              Lippard@RVAX.CCIT.ARIZONA.EDU
Dept. of Philosophy      Lippard@ARIZRVAX.BITNET
University of Arizona
Tucson, AZ 85721

From:    Jan Willem Nienhuys
Subject: Re: sTARBABY to death

>The effect is only supposed to work with natural births, not drug-induced.
>That's why he only collected births before a certain year.  Prior to the
>CSICOP test of American athletes, Gauquelin said that he hadn't found the
>effect for basketball players, and so they shouldn't be used in the test.
>They were used anyway.  I don't believe Gauquelin ever placed any
>restrictions

More precisely, G. said *after* the CSICOP test that he could have told
that basketball players were unsuitable, because he had bad results with
Italian basketball players.
If one examines the sportsmen he included in his own 2088 sportsmen, you
will find there enormous amounts of Italian sportspilots (many of them
military), you will find quite a few Basque pelote players (pelote is
a kind of ball game only played by Basques) but no (repeat: no)
basketball players.  So he decided to leave them out after he had seen
their Mars sectors.
Now if you know that just about 15 less `Mars' sportsmen would have
made the difference between significant or not (in the physics sense,
namely with a 3-sigma level), and if you know that after the CSICOP
test (which wasn't perfect) he indulged in an quite a lot of post-hoc
sample splitting and claiming that only just about Olympic Gold Medallists
are good enough (his own samples having no compunction about several
instances of two or three brothers), than it is not hard to
understand why a certain amount of skepticism seems justified regarding
G.'s data gathering technique. It need not have been intentional, but
more a case of `making up the hypotheses as you go', but then those same
2088 sportsmen should not have been presented as proof.
This leaves exactly two tests of G.'s hypothesis: one with the Belgians,
one with the Americans, one positive, one negative.