30

Ratios of 0.25 to 4.00

99

98

20

95

90

70

10

50

30

0

10

5

2

1

0.1

1.0

5.0

QC1

QC

Logs of

Concentration Ratios

TPH

Concentration Ratios

QC2

QA

these ratios on lognormal paper (Fig. 9). How-

0.224.07, which is in good agreement with the

ever, the histogram of ratios for MRD results

arbitrary limits used to edit the data. The geo-

shows extreme dispersion, and the values between

metric mean of the 82 QC/QA ratios between

0.254.00 form a rectangular distribution, thereby

0.254.00 was 0.98 with 95% confidence limits of

precluding statistical analysis using a lognormal

0.841.15 and 99% tolerance intervals of 0.166.15.

model. The NPD results were probably more reli-

able than the MRD results because the former

99

tried to ensure that both QC and QA laboratories

98

used the same state-approved method. There is a

95

wide variation in both the extraction and analysis

procedures for TPH as a function of various state

90

requirements. We chose to analyze only the more

reliable NPD results; no further analysis was con-

70

ducted on the MRD results. This decision was

rooted in the conviction that future aspirations

50

should at least be based on the best of past per-

formances.

30

A histogram of the log QC1/QC2 results is

shown in Figure 10 and the lognormal probabil-

10

ity plot appears in Figure 11. Both plots provide

5

only fair fits to the lognormal model and there is

some uncertainty about the correct location of the

2

straight line in Figure 11. Nonetheless, the expec-

1

tation of lognormality and the absence of a better

0.1

1.0

4.0

model convinced us to proceed with this model.

QC1

TPH

Concentration Ratios

QC2

The geometric mean of the 78 QC1/QC2 ratios

between 0.254.00 was 0.95 and 95% confidence

limits on this mean were 0.841.07 (Table B5). The

99% tolerance intervals for individual ratios were

13