The “Bad French” Fallacy
LNDD officials last week asserted that the leaking of embarrassing documents to various journalists and anti-doping agencies was the work of a hacker, because this person used poor grammar and mis-spelled the name of the town where the lab is located when writing cover letters to the recipients. Perhaps.
But, then again, one needs to ask:
- Do the lab technicians need to pass a grammar test in order to begin employment at LNDD?
- And for that matter, are all of the lab technicians native French-speakers?
I’ve encountered a lot of writing by highly educated English-speakers that contains grammatical and typographical errors, but I wouldn’t look at their writing and say that it was the product of a non-native. I don’t know that the LNDD justification for believing an English-speaking hacker got hold of their documents holds water.
Given the nature of the documents released, I suspect that it was someone with knowledge of the lab and the cases it’s handled who released the documents. Whether that’s a current or former employee, I don’t know. But it sure seems like an inside job.
Gotta Check Your Facts Department
My bad. In yesterday’s post, I spoke of an extreme variance in the T/E results for Floyd Landis’ A sample. Part of the explanation for the variation is that two different tests were performed. A quick test to determine whether the T/E ratio was above the threshold, and a lengthier, supposedly more accurate, test to confirm whether the threshold had been exceeded.
The real question that should be asked about the results is why the two different tests yield such different results. If the difference between the two tests is consistent, then the “more accurate” test would always find a T/E ratio more than twice the value of the screening test. So, for example, a person with a T/E ratio of 2 on the screening test would have a T/E ratio of greater than 4 on the confirmation test.
On the surface, that doesn’t make sense, because in such a scenario a large number of potential violations would be missed. That’s assuming you accept that a T/E ratio of 4 is a good threshold value. Arnie Baker makes some good points about why the threshold should be 6 or even higher in his Slide Show 2.0.
That said, I’m rethinking whether the variance in the A sample T/E results is something Landis’ defense team should even argue about. The contamination of the B sample seems a stronger point of contention, and I think they should probably concentrate on that. The lab’s own data showing the contamination will be harder for LNDD, WADA and the UCI to refute.
Did Landis’ A Sample Become Contaminated?
Just a bit of speculation on my part, but could the explanation for the variation in the T/E results between the initial screening test and the confirmation test be due to contamination? The initial screening was conducted on 7/21 and the confirmation on 7/24. If the sample spoiled due to improper storage, could that account for the results? Or is it there something about the two different testing methods that accounts for the discrepancies?
As we’ve already seen, the B sample exceeded WADA’s threshold for contamination. Perhaps the A sample did, too. If not at first, perhaps during the course of the testing process.
Changing The Rules Department
In today’s Cycling News, there’s an article that the Italian association of cycling medics (AIMEC) are changing some of the rules and procedures for anti-doping testing in 2007.
Starting next year, athletes who will be tested following the race will be met at the finish line and escorted to the doping control area in order to prevent riders from “preparing themselves” for the test in between being notified (usually late in a race) and arriving at the doping control area.
They are also considering conducting random blood tests up to 30 minutes prior to the start of an event. More of a concern is this:
[Scientific consultant to the UCI Mario] Zorzoli announced that starting in 2007 there will be a lowering of the allowed testosterone level (following on the positive test by Landis at the Tour de France)
As Arnie Baker showed in his slide show, of about 1000 athletes who were followed up after an initial T/E ratio between 4.0 and 6.0, only 3 were subsequently confirmed as being positive for doping.
What’s not clear from the article is whether they mean the actual amount of testosterone present, or whether they mean a change in the T/E threshold value to conduct follow-up tests.
Regardless, this makes me concerned that even more innocent people are going to be suffering the same fate that Floyd currently is, unless something within the system changes to protect the athlete’s privacy before the process runs its course. Too many more of these trial-by-media scandals and we may see permanent damage to the sport, or worse — the complete collapse of professional competitive cycling.
Dear Rant:
I was discussing the dispersion of t and e in the aliquots with my husband, who is a lawyer for the FDA. Now, he just hears about these issues from scientists and statisticians, so no one take this as gospel, but he tells me it is possible for the concentration of a substance in an aliquot to vary a great deal if the aliquot is very small. Nature does not allow an even dispersal — consider stars in the sky. They aren’t distributed evenly, and if you look at a small patch, you may see none.
He suggests that the extensive range means the aliquots are so small there is more chance of uneven dispersion and therefore error in the result. The tricky part is correcting for the error in such a way that the average can be with 95% certainty be considered correct (the DPF discussed the 95% a lot and I have the utmost respect for those who could follow the discussion as I got lost there).
The other tricky part, and which could support Mr. Landis assuming the DPF hasn’t gone over this issue with a fine tooth comb and dismantled it, is how much of the range was under the appropriate threshold for an adverse result. So for example, if the threshold for an adverse reaction is 4, the average of the test is 5, and the average results from 4 aliquots of 11, 1, 4 and 4, then you have 3 results under the threshold but an average that creates an adverse result. My spouse feels many labs would discount this result as too many numbers are below the threshold and the higher number could be an outlier.
We are both appalled that WADA hasn’t seen fit to have standard protocol for reading results throughout the labs — another area of inconsistent results from an organization whose purported purpose is standards of fair play.
Cheryl,
Busy day here in the salt mine. Barely had a chance to read Floyd’s comments, quoted over at TBV. You’ve given me a good idea for a next article (just how certain do you need to be that the data is accurate). I need to do a bit of fact-checking before I write it up. Perhaps tonight. Or if not then, tomorrow.
– Rant
Dear Rant:
Please fact check a lot (I’m guessing the 2nd “you” means me and not you — pronouns are tricky). I am an art historian. My husband is an attorney. We are not scientific reliable sources. I think of Marc over at TBV who had some great ideas, but then the scientists at DPF told him what’s what, and he has homework facing him.
I believe I saw the bit about lack of consistent WADA lab standards for a adverse result somewhere on TBV. Tom Fine who runs the LandisWiki has my vote (for what it’s worth – can’t you tell I’m caveating all over the place) as the man for science info.
Cheryl
Cheryl,
Actually, what I meant to say was: Just how certain does a test’s accuracy need to be? Should it be 95% accurate, 99% or …?
That’s the direction I’m going in for the evening.
– Rant