Figure 6. Lognormal probability plots for QC/QA concentration ratios for soil VOCs

SR96_09
Page Navigation
10

Figure 6. Lognormal probability plots for QC/QA concentration ratios for soil VOCs.

likely we are to experience biases and the more

and, therefore, they omit all error associated with

difficult it becomes to identify them.

sample collection, transport, storage, and extrac-

What about the QC results from matrix spike/

tion. Assuming that the solution used for spiking

matrix spike duplicates (MS/MSD)? We could find

is prepared independently from the calibration

no meaningful correlation between these results

standards (frequently untrue), MS/MSD results

and the degree of agreement between QC and

can serve to validate calibration and to reveal spe-

QA laboratories. Similar failure of MS/MSD re-

cific interferences from the matrix under analysis.

sults as a quality indicator for VOCs was recently

Does this imply that calibration variations do not

reported by Robertson et al. (1995). Generally

contribute to the disagreement between laborato-

speaking, the MS/MSD results appeared to be

ries? The answer is no. Significant bias may be

quite satisfactory and suggestive of much greater

introduced from calibration errors, but we do not

reliability than was actually obtained on co-lo-

believe it is the major source of error. In contrast

cated samples. However, we must remember that

to the soil results, we found that most QC/QA

MS/MSD results are derived by spiking extracts

ratios for GC-MS determination of VOCs in