La. Currie, Some case studies of skewed (and other ab-normal) data distributions arising in low-level environmental research, FRESEN J AN, 370(6), 2001, pp. 705-718
Three general classes of skewed data distributions have been encountered in
research on background radiation, chemical and radiochemical blanks, and l
ow levels of Kr-85 and C-14 in the atmosphere and the cryosphere. The first
class of skewed data can be considered to be theoretically, or fundamental
ly skewed. It is typified by the exponential distribution of inter-arrival
times for nuclear counting events for a Poisson process. As part of a study
of the nature of low-level (anti-coincidence) Geiger-Muller counter backgr
ound radiation, tests were performed on the Poisson distribution of counts,
the uniform distribution of arrival times, and the exponential distributio
n of inter-arrival times. The real laboratory system, of course, failed the
(inter-arrival time) test - for very interesting reasons, linked to the ph
ysics of the measurement process. The second, computationally skewed, class
relates to skewness induced by non-linear transformations. It is illustrat
ed by non-linear concentration estimates from inverse calibration, and biva
riate blank corrections for low-level C-14-C-12 aerosol data that led to hi
ghly asymmetric uncertainty intervals for the biomass carbon contribution t
o urban "soot". The third, environmentally skewed, data class relates to a
universal problem for the detection of excursions above blank or baseline l
evels: namely, the widespread occurrence of ab-normal distributions of envi
ronmental and laboratory blanks. This is illustrated by the search for fund
amental factors that lurk behind skewed frequency distributions of sulfur l
aboratory blanks and Kr-85 environmental baselines, and the application of
robust statistical procedures for reliable detection decisions in the face
of skewed isotopic carbon procedural blanks with few degrees of freedom.