there's a potentially interesting statistical question in relation to OPCW lab results. Analyses were done at two labs (DL02,DL03). Other than a few chloride values, only presence/absence was reported. A surprising degree of inconsistency between labs.
2/ for example, 16 samples were reported from the critical Location 2 (cylinder on balcony location), yielding 32 analyses for chlorophenol (16x2 labs). 10 samples null for both labs, 1 sample positive for both labs, 5 samples positive in only one lab ( DL03-3; DL02-2). 
3/ in order for this pattern to arise, AT LEAST 5 (of 32) analyses are wrong (one analysis from each of the 5 inconsistent results.) I.e. there is an error rate of approximately 5/32= 15.625%.
4/ applied to 16 samples, a 15.625% error is most likely going to yield 2-3 false positives (by simple application of binomial theorem from high school).
5/ only one Location 2 sample had chlorophenol in both samples: 25SDS - wood fragment from kitchen door on "level 2" (3rd floor).
6/ the other three positive chlorophenol analyses for Location 2 by lab DL02 are all from concrete debris, a material, which, unlike wood, is not a possible precursor to chlorophenol. Note co-presence of TNT in these samples.
7/ on an empirical basis, I've noticed co-presence of TNT and chlorophenol in an anomalous number of samples, something that I'll look at.
8/ lab testing for chlorophenol appears to be extremely accurate, detecting values down to 2.5 ppb. The only plausible explanation for inconsistency between lab results for so many samples (5 of 16) that I can think of is that these 5 samples are very close to detection limit.
9/ if these 5 samples are at detection limit (say ~2-3 ppb), that could explain why they are detected by one lab but not the other. Since each lab detected values not detected by other lab, bias in one lab would not explain this.
