in Greenland ice core ” by N.J.G. Pearce et al.

• Main Author: Douglas J. Keenan
• Other Authors: The Pennsylvania State University CiteSeerX Archives
• Format: Text
• Language: English
• Published: 2004
• Subjects: Greenland; Greenland ice core; greenlandic; ice core; Alaska
• Online Access: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.488.308; http://www.informath.org/G%5E304aS.pdf

Pearce et al. [2004] compared tephra that was retrieved from Greenlandic ice (by Hammer et al. [2003]) with tephra from Aniakchak, Alaska. Largely on the basis of statistical analyses of the geochemical data, Pearce et al. claimed that the Greenlandic tephra matches with Aniakchak. Herein, those statistical analyses are examined. Pearce et al. [2004] devote a section of their paper to discussing the “statistical distance ” between two sets of tephrochronological data. The formula for statistical distance involves standard deviations. The term “standard deviation”, though, can mean different things in different contexts: either the potential error of a measurement (i.e. precision) or the inter-particle dispersion (i.e. indicating variation among different particles). Pearce et al. [2004, table 2, caption] note this distinction when presenting their data. In applying the formula for statistical distance, however, they confuse the two meanings: the formula requires measurement precision, but they use inter-particle dispersion instead. This confusion makes the statistical-distance calculations of Pearce et al. [2004, sect. 6] erroneous. (There is no immediate way of fixing the calculations, because when the correct meaning of “standard deviation ” is used, the D2 statistic of Pearce et al. no longer has a chi-squared distribution.) A second statistical problem with the work of Pearce et al. [2004] is in the use of “relative differences”. The problem is easily seen via a hypothetical example: consider two tephras, both measured by two laboratories, with the following results for SiO2 (on bulk samples, with measurement standard deviations indicated).