Table 9 (cont'd).
b. Statistical analysis of TNT concentrations (g/g) for discrete
and composite samples.
Discrete samples
On-site analysis
Laboratory total
Sample
Mean
Mean of logs
Mean
Mean of logs
4.2f†
1
0.625g
7.2d
0.854d
2
6.0ef
0.779f
7.6d
0.879d
3
18.7b
1.271b
24.3b
1.385b
4
10.6d
1.024d
8.8d
0.945d
5
34.4a
1.536a
40.5a
1.607a
6
14.1c
1.151c
14.7c
1.167c
7
7.6e
0.879e
8.7d
0.940d
†
Numbers designated with the same letter are not significantly different at the
95% confidence level.
ANOVA for on-site and lab analyses
Untransformed
On-site
Lab
F ratios
217***
321***
Error MS
0.9974
95,854
Least sign. diff.
2.36
2.32
Analysis s
1.0
1.0
Sampling s
10.4
12.4
(s = standard deviation)
Linear correlation analysis for on-site analysis vs. lab analysis
(r = correlation coefficient)
Slope
Intercept
r
untransformed, non-zero intercept
0.990
0.856
0.984
untransformed, zero intercept
1.032
0
0.982
log-transformed data
1.000
0.019
0.939
Results of paired t-tests for on-site vs. lab results
Means of seven discrete samples, t = 2.17 (NS)
Composite samples
On-site analysis
Laboratory total
n
7
7
mean value
16.6
14.9
standard deviation
1.52
0.33
RSD
9.0%
2.8%
ANOVA comparing on-site and lab analyses
F ratio = 8.43*
* Significant at the 95% level
*** Significant at the 99.9% level
** Significant at the 99% level
NS Not significant at the 95% level
Correlation analysis with the field and lab data
were significantly different from one another.
from location 9 gave a best fit linear relationship
When variances were fractionated into analytical
with a slope 0.990, a y-intercept of 0.856, and an r
and sampling error, a standard deviation for analy-
sis of 1.0 g/g was obtained for both the field and
of 0.984 (Table 9b, Fig. 7). The best fit zero inter-
cept model had a slope of 1.032 and an r of 0.982,
laboratory methods. Sampling error estimated
from the field analysis data was 10.4 g/g and
suggesting that the intercept was probably not
from the lab data it was 12.4 g/g, showing that
significant. Results of a paired t-test indicated
that the results of the two methods were not sig-
sampling error again dominated the total error at
nificantly different at the 95% confidence level.
this sampling location (Table 9b).
31
Previous Page
