the same lines in the gamma-ray spectrum. The results for 232Th and 228Ra should be comparable because both are actually based on measuring gamma-rays from the decay product 228Ac. The
results for rays from
238 234 U and Th.
234Th should be comparable because both are actually based on the gamma-
For both of these statistical tests, each sample for which one or both labs failed to detect and measure the radionuclide of interest was omitted from the data set.
Test 1: For the first test, each difference was normalized by dividing it by its standard uncertainty (i.e., an estimated standard deviation).5 A two-sided t-test was then performed on the normalized differences to test whether the mean normalized difference was different from zero at the 5 % level of significance. For 226Ra, 40K, and 238U/234Th no significant differences were detected. However, a statistically significant difference was found for 232Th/228Ra. NAREL’s
results for 228Ra were significantly higher than FDOH’s results for based on the activity of the gamma-emitting decay product 228Ac).
232Th (both of which were
Test 2: For the second test, the differences were not normalized. A two-sided t-test was performed on the unnormalized differences to test whether the mean difference was different from zero at the 5 % level of significance. Test 2 produced the same conclusions as Test 1 for all four radionuclides. A statistically significant difference (at the 5 % level) was found only for
The “relative percent differences” (RPD) between NAREL’s results and FDOH’s results were also calculated and averaged by nuclide. The RPD is defined as follows
N–F (N + F)/2
× 100 %
where N = NAREL’s result for a particular sample and F = FDOH’s result for the same sample. Although the sample-by-sample RPD values vary from less than 100 % to more than +100 %, the average RPD is large (37 %) only for 232Th/228Ra.
228 Th/ Ra
234 U/ Th
226 40 Nuclide Ra K
RPD +4.2 %
0.5 % +37.0 % +6.8 %
Note: Some of the normalized differences for particular samples and nuclides are unusually large. If there were no systematic differences between the labs and if the measurement uncertainties were realistic, one would expect nearly all the normalized differences to fall between 3 and +3. In fact several of the normalized differences exceed ±3 and one of them exceeds +5. The large difference between the two 40K results for sample 90675-04963 is particularly hard to explain, since 40K is generally considered to be easy to measure. NAREL double-checked its result for this sample, which is similar to the 40K results for other samples. FDOH’s result for this sample is smaller than the 40K result for any other sample in the group.
5 The uncertainties reported with the FDOH results represent counting uncertainty only. The uncertainties reported by NAREL include counting uncertainty and other uncertainty components.