On the Correct Use of Statistical Tests: Comment on “Lies, Damned Lies, and Statistics (in Geology)”

Authors

  • D. Sornette,

    1. Department of Management, Technology and Economics, Eidgenössische Technische Hochschule (ETH) Zürich,Zurich, Switzerland
    2. Department of Earth Sciences, Eidgenössische Technische Hochschule (ETH) Zürich,Zurich, Switzerland
    Search for more papers by this author
  • V. F. Pisarenko

    1. International Institute of Earthquake Prediction Theory and Mathematical Geophysics, Russian Academy of Sciences,Moscow, Russia
    Search for more papers by this author

Abstract

[1] Taking the distribution of global seismicity over weekdays as an illustration, Pieter Vermeesch (Eos, 90(47), 443, doi:10:1029/2009EO470004, 2009) in his Forum presented an argument in which a standard chi-square test is found to be so sensitively dependent on the sample size that probabilities of earthquake occurrence from these tests are uninterpretable. He suggests that statistical tests used in the geosciences to “make deductions more ‘objective’” are at best useless, if not misleading. In complete contradiction, we affirm that statistical tests, if they are used properly, are always informative. Vermeesch's error is to assume that it is possible to reduce in the chi-square test simultaneously both the number of earthquakes in each weekday and the sample size by 10. Instead, Vermeesch should have taken 10% of the original data set and then again grouped it into 7 days. Without doing this, it was inevitable that Vermeesch reached his erroneous conclusion.

Ancillary