Plutonium - 239,240. There is an outlier (Lab C) in the Batch 2 results (too high by a factor of 2) and
this value was excluded in computing the mean. Besides the difference in magnitude, this result can
also be discarded based onits calculated 239,240py- to -241 Am ratio of 4.5, which is far above the
Known ratio of 2.3 + 0.4 for that area of Janet. With that number deleted, the 99 percent confidence
interval on the mean of Batches 1 and 2 contains all but the Lab B Batch 1 result of 77.2 pCi/g. This
value is 15 percent higher than the mean, but is only 0.3 pCi/g higher than the upper limit of the
confidence interval. All the results for all labs are within the 99 percent confidence interval on the
mean of Batches 3, 4 and 5. Therefore, except for the one outlier, interlaboratory agreement is good
for these isotopes.
Cesium - 137. Results for all laboratories are within the 99 percent confidence interval on the mean
of Batches | and 2, and all but one are within the 99 percent confidence interval on the mean of
Batches 3, 4 and 5. The exception is the Lab B Batch 5 value, which is 11 percent higher than the
mean, but is only 0.3 pCi/g higher than the upper limit of the confidence interval.
interlaboratory agreement is good for this isotope.
Thus
Strontium - 90. There were some problems noted in the Batch 1 results for this isotope, and at the
time it was unclear which of the disparate results was more accurate. The Batch 2 results indicated
the Lab B and D results for Batch 1 might not be reliable. The 99 percent confidence interval on the
mean of Batehes 1 and 2, computed with those two samples eliminated, contains all but those two
samples. Including those samples more than doubles the standard deviation, leading to the conclusion
that the Batch 2 results for Lab B are reliable, but the Batch 1] results are not.
There was also a problem in Batches 3, 4 and 5. The Lab A resuit is an outlier, while all other results
lie within the 99 percent confidence interval on the mean, computed with the outlier excluded.
Fortunately, Lab Al conducted the analysis for the 90s- data actually used and it shows good
agreement with other labs.
Conclusions. Overall, agreement among laboratories was good. These comparisons indicate that the
results from Lab A (the EIC Enewetak laboratory) were reliable with the exception of 90sr. This
caused no severe problem since Lab Al (the EIC Albuquerque laboratory) provided the 90s- data used
for the dose assessment and Lab Al results were supported by Lab B forthis isotope.
5.2.5 Other Programs
Statistical design and analysis were required for several programs and experiments not directly
related to the surface and subsurface soil cleanup efforts. Among these was the plowing experiment
(Plow-X) that was an investigation of a possible alternative or adjunct to surface soil removal. The
experiment was designed to check the effects of deep plowing on both surface TRU activity and
distribution of activity as a function of depth. The surface comparison used a randomized block
design and data from the IMP. The subsurface investigation involved a multivariate analysis of
variance on soil profile data. The soil samples were taken in a pattern that was selected to avoid
confounding the effects of plowing with effects from using a backhoe to dig the sampling trenches.
Details on the experiment and the philosophy behind it are in Section 6.7, and the results of the
statistical analyses are in Tech Note 9.1.
Comparisons of 241 4m data from the IMP with laboratory 241 4m results from surface soil samples
also involved statistical analyses.
The earliest work, using a regression approach on Janet data
(Barnes, 1978), resulted in the conclusion that the two types of data agreed reasonably well.
Continuing questions about the accuracy of the IMP data, however, prompted more analyses using a
somewhat different regression method and data from several islands. There were some significant
differences between the two data types, so an investigation was made of the variability of 241 am
activity in soil.
A description of the investigation and results are reported in Teeh Note 8.0.
The
results indicated that statistical investigation of the possible differences between soil and IMP data
would always be difficult because of the high variability of 241 am activity in soil.
Theoretical calculations eventually led to discovery of a bias in the in situ data due to incorrect
assumptions of the soil composition, density and moisture parameters used in deriving the IMP
conversion factor. Tech Notes 22 and 23 describe the collection of additional data to arrive at more
accurate parameters and the final correction, respectively. (See also Sections 3.2.6 and 3.2.9.)
146