In addition there is a REECo internal quality control procedure.

For

instance, the analytical balance in the soil preparation area has a quality
contro! program going on it.

There will be resampling of a large batch of

soil recent ly taken at Hurricane and both plutonium and gamma analysis samples will be sent through on a periodic basis.

Given that the plutonium

analysis: can handle eight per week, eight samples per week, one of those

will

be an external
=

a

quality control

sample,

and one of them will be an

internal quality‘control sample.
We have’‘designed an experiment to address a problem in precision of
10

analysis in the REECo laboratory.

1l

analysis of a randont=components design and should allow us to determine at

12

what part

13

creeping in.

14

many people care

15

crowd, this is what R. L. Anderson calls a staggered experimental design.

16

We will take eight 200-gram al iquotsfour of this kind, two of this kind,

17

one of each of those kinds.

18

the sample.

19

two analyses here will give us an est imateofthe variability due to the

20

plating step.

21

estimate.

22

and this will give us, directly, an aliquoting estimate, or have 1 skipped

23

one, leaching.

24

degrees

25

Type 2, Type 3 and Type 4, and what we have at the end of the experiment is

26

‘an almost equal number of degrees of freedom for estimating each of the

in the REECo plutonium analysis procedure,

variance

28

design

the uncertainty is

Perhaps F-coutd stick this up here (FM-4), probably not too
too much

about-this,

but

for

the

statisticians

in

the

Everytime there is a fork in there we split

You notice that we split one of the samples 16 times and the

The split here will give us -acolumn-extraction variability

The split here will give us a scavenging-variability estimate,

I have laid out here, given that this—is Type 1, where the

of freedom with respect

27

The experiment, essentially, is a basic

components.

An

to the exper imentatdesign

additional

nice

property of

this

go

and

for

experimental

is that the estimates of variance components are not correlated,

175

Select target paragraph3