James J. Filliben
Keith R. Eberhardt
Mark G. Vangel
Eric S. Lagergren
Statistical Engineering Division, CAML
Ionizing Radiation Division, PL
In today's pollution-sensitive world, it has become increasingly more important to have available well-characterized standard reference materials of a natural "matrix" type which may be used in laboratory tests of measurements of environmental radioactivity. Such materials would be used in the evaluation of competing analytical methods, and and also in the cross-comparison of interlaboratory data-both at the national level and the international level.
Ken Inn and Zhichao Lin of the Ionizing Radiation Division of the Physics Laboratory have constructed such an SRM (4357)-consisting of a 12-component natural matrix ocean specimen. The 1000-bottle SRM 4357 will have two sediment sources: the Chesapeake Bay (benign) and the Irish Sea ("hot"). The project goal is to determine global (valid across all 1000 bottles) values and uncertainties for each of the 12 elements (radionuclides): potassium, radium, thorium, strontium, uranium, etc.
The certification required three steps-all with technical challenges: 1) creating a natural matrix specimen starting with raw sediments (careful specimen blending, diluting, pulverizing, re-blending, sterilizing and distributing into the 1000 bottles);
2) physical fractionation of the specimen into elemental components (a determination of optimal procedural settings for signal maximization -this will require future design of experiment consultation);
3) determination of certified values and uncertainties for each of the 12 elements (a detailed data analysis of the interlaboratory data from 18 international labs, and for each of the 12 certification & uncertainty values, involving 1) interlab comparison; 2) outlier rejection; 3) distributional analysis; 4) homogeneity testing; 5) homoscedasticity testing.
An interesting distributional conclusion was that the 8 ``natural" radionuclides tended to have a normal distribution, while the last 4 ``man-made" radionuclides tended to have a Weibull distribution.
Note that this project collaboration was a direct result of the Spring 1995 Design of Experiment Course (Filliben & Lagergren) for NIST Scientists & Engineers in which one of the authors (Zhichao Lin) was a participant.
Figure 3: Environmental SRM Distributional Analysis. Bootstrap estimation of the 3-parameter Weibull shape parameter and the corresponding 2.5 percent point
Date created: 7/20/2001