Table 5. Summary of edited results for VOCs in the three soils.
Soil A,
TDCE
TCE
Ben
Tol
Tampa Bay sediments
A1
A2
A1
A2
A1
A2
A1
A2
Batch means (g/g)
6.73
6.04
8.84
8.32
7.56
6.92
10.5
9.93
Overall mean* (g/g)
6.39
8.58
7.24
10.2
Std. dev. (g/g)
1.80
0.85
0.71
0.95
RSD (%)†
28.2
10.0
9.8
9.3
95%CL**(g/g)
5.57.3
8.29.0
6.87.6
9.710.7
Soil B,
TDCE
TCE
Ben
Tol
Rocky Mtn. Arsenal
B1
B2
B1
B2
B1
B2
B1
B2
Batch means (g/g)
11.0
11.1
15.5
15.6
14.1
14.1
19.7
19.0
Overall mean* (g/g)
11.0
15.6
14.1
19.4
Std. dev. (g/g)
1.9
1.7
1.2
1.8
RSD (%)†
17.0
10.8
8.5
9.5
95%CL**(g/g)
10.111.9
14.816.4
13.514.7
18.520.3
Soil C,
TDCE
TCE
Ben
Tol
Pt. Barrow, Alaska
C1
C2
C1
C2
C1
C2
C1
C2
Batch means (g/g)
39.1
39.9
55.7
56.7
39.2
39.8
63.6
63.4
Overall mean* (g/g)
39.5
56.2
39.5
63.5
Std. dev. (g/g)
5.4
7.6
3.5
7.7
RSD (%)†
13.6
13.5
8.8
12.2
95%CL** (g/g)
37.042.0
52.659.8
37.941.1
59.967.1
* There were no significant differences in batch means (α = 0.05); batches were combined.
† Relative standard deviation.
** 95% Confidence limits on overall means.
One important conclusion from these results is
Steiner 1975). The ranges were homogeneous ex-
that none of the batch means differed significantly
cept for the laboratory 16 results for soil B, where
from its replicate (α = 0.05). In fact, only Ben in
three of the four analytes were outside of accept-
able limits (α = 0.05). Since subsample B2 had
soil A yielded a t statistic that even approached
the critical value for significance. This finding sug-
been flagged earlier due to a very low recovery of
gests that batches of subsamples can be reproduc-
the internal standard, soil B results for laboratory
ibly prepared over time using a given soil and a
16 were excluded.
given fortifying solution of VOCs. Further evi-
dence for the predictability of the procedure is
Edited data set
At the conclusion of these tests, we were left
seen in Table 6, where analyte percentages rela-
with complete data sets for nine laboratories for
tive to the total of the four analytes are computed
soils A and B and ten laboratories for soil C. Labo-
for each soil. The percentages are quite consistent
ratories 3, 12 and 13 were excluded for all soils,
despite substantial differences in the clay and or-
laboratory 5 was excluded for soil A, and labora-
ganic matter content of the soils, which suggests
tory 16 was excluded for soil B.
The ANOVA confirmed that significant labora-
Table 6. Percentages of each
tory differences remained (α = 0.05), along with a
spiked analyte relative to the to-
tal of the four analytes in each
small but statistically significant interaction be-
soil.
tween laboratories and samples. This situation
was not unexpected in view of our earlier decision
Percent of each analyte
to retain two laboratories with significant system-
relative to the total
atic errors based on the ranking test. We believe
Analyte Soil A Soil B Soil C Average
that these small systematic errors are detectable
TDCE
19.7
18.4
19.9
19.3
because of the excellent precision obtained. Fur-
TCE
26.5
25.9
28.3
26.9
thermore, we feel that no further exclusions are
Ben
22.3
23.4
19.9
21.9
necessary or desirable. A summary of the edited
Tol
31.5
32.3
31.9
31.9
Total
100.0
100.0
100.0
100.0
results appears in Table 5.
6
Previous Page
