W. J. Bair
June 18, 1979
Page 2
task but in view of the scrutiny this dose assessment will receive,
I feel LLL needs this kind of documentation to lay out the available
information.
Desert Research Institute statisticians at Las Vegas
would seem to be the logical choices for spearheading this project.
However, their work should be reviewed by statisticians outside the
DOE community as well as by ourselves (Advisory Group).
General Comments on Working Draft
As acknowledged by Robison, the present preliminary manuscript is incomplete
in a number of important areas.
The main areas of deficiency from my statistical
point of view are:
1)
Inadequate description of how the diet survey was conducted and how
the resulting data were statistically analyzed.
Since the dose
estimates are based on these diet data, it is crucial that the
magnitude of sampling errors by quantified.
It is discouraging,
to say the least, that the authors did not indicate the number of
people interviewed in the survey.
Any statistician (or lawyer)
worth his salt would choke on that omission.
Also, no details are
given on how the survey was actually conducted.
This methodology
must be known before we can evaluate whether the survey data are
more reliable than expert opinion.
2)
Completely inadequate description of how the soil data were
statistically analyzed and summarized in preparation for use in the
dose models.
For example, we do not know how "Jess-than" concentra-
tions were handled statistically.
It is my understanding that on
Engebi there were few "less than" concentrations obtained.
However,
on other islands most of the data at depth were less than detection
limits.
I suggested to W. A. Phillips several weeks ago that they
compute the dose estimates using average soil concentrations computed
in two ways:
(1) using the MDA (minimum detectable activity) for
each less-than number, and (2) using the actual concentration reported
whether it be negative, positive or zero.
This would indicate how
much dose estimates might differ depending on how these “less-than”
data are treated.