Astrology and statistics
Gauquelin 'n' co

I'm studying the history of published statistical tests on astrology between 1955 to 1996, which led me to develop opengauquelin.org, a database gathering the data used in the historical tests. This project is detailed on tig12.github.io/g5.
My question is : Can we observe irrational phenomenons using only rational means ?
Current question has been reduced to : Can we detect statistical anomalies related to astrology without human bias ?
From there, an other question arises : Can we observe selection bias in the historical tests ?
Studying this story convinced me that this problem has not been studied seriously by the academic world, except by two German psychologists, Suibert Ertel and Arno Müller, who analyzed all the published tests and conducted independant tests. Their conclusions are :
  • Introducing an "objective" method to select the most eminent persons, the "mars effect" can be observed, and the effect is stronger for very eminent athletes.
  • The reproduction of previous tests and independant replications sometimes confirm the existence of anomalies, and sometimes show that claimed anomalies can't be reproduced.
  • Selection bias from Gauquelin during the "Comité Para" test (Belgium, 1976) - this test showed the existence of statistical anomaly.
  • Manipulation of data by Paul Kurtz in the CSICOP test (USA 1979) to make the anomaly disappear.
  • Opaque and unfair data selection in the CFEPP test (France 1996) makes the conclusion of this test (inexistence of anomaly) not scientifically valid.
Conclusion : The question is still open.
I believe that it is today possible to re-examine these tests and build a more precise knowledge about these questions. To reach this goal, the central point is data : build more complete and reliable groups of famous persons and sort them by eminence.