36

Oil  Spill  identification  -  Round  Robin  20045

Table  3.
Normalizing  the  peak  heights  on  peak
4-M-pyrane  (e)  of  mass  216  (LVA)

Fig  12
Ratio  plot  of  Cedre  including  error  bars
showing  a  95%  interval  calculated
from

Another  example  is  the  ratio  calculation  of  LVA  shown  in  Table  3.  The
table  highlights  the  difference  in  ratio  between  B(a)F/  4-methylpyrene
for  Source  II.

Ratio

EXTR  I

EXTR  II

Source  I

Source  II

a/e

0.23

0.23

0.25

0.16

b/e

1.0

0.94

0.93

0.36

c/e

0.19

0.16

0.16

0.1

d/e

0.79

0.76

0.8

0.72

f/e

0.73

0.70

0.75

0.60

Such  tables  show  differences,  but  what  is  acceptable  is  still  a  personal
decision.
In  the  CEN  guideline  it  is  defined  that  differences  caused  by  the
analytical  method  may  not  lead  to  the  conclusion  that  samples  are
different:
positive  match
differences  in  chromatographic  patterns  and  diagnostic  ratios  of  the  samples
submitted  for  comparison  are  lower  than  the  analytical  variance  of  the  method
or  can  clearly  be  explained  by  weathering.  The  samples  are  identical  beyond
reasonable  doubt
Several  labs  have  used  numerical  limitations  to  estimate  the  acceptable
difference.
70
60
50
40
30
20
10
0
Figure  5  Aromatics  distribution  of  4  samples  (in/z  =  216  and  234)  (error  bars  represent  95%
confidence  level  intervals,  N=3)

m  Source  I  ■  Source  II  □  Extract  I  □  Extract  II

B(a)F/4-Mpy  B(b+c)F/4-Mpy  2-Mpy/4-Mpy  1-Mpy/4-Mpy  Retene/C4-Phe

Fig  12  shows  a  bar  graph  of  Cedre  with  error  bars:
The  analyses  were  run  in  triplicates  for  each  sample,  and  error  bars
considering  95%  confidence  level  intervals  were  plotted  on  the  histogram.
In  the  report  of  Cedre  also  graphs  with  a  95%  and  98%  confidence  are
shown  and  used  to  estimate,  whether  differences  are  significant  or  not.