tafnut wrote:Obvious (i.e., stupid) question on my part:

What does this do to the credibility of WADA/UADA/IAAF/USATF?

How can anyone take anything seriously anymore?

a. A 'negative' test does not necessarily mean the athlete did NOT use a PED. He just didn't get caught.

b. A 'positive' test does not necessarily mean the athlete DID use a PED. The lab made an 'error'. [to what DO they ascribe Marion's +/- results?]

Good luck on this, ABC Boiz.

You need better background on testing in a "statistical world". You want to both have a low likelihood of having a false positive and of having a false negative. One way to do this is to increase the sample size, which gets us to the A/B stuff. Another is the set of test charactreistics. A number of tests have large "regions" where the test is very likely + or -. This arises, for instance, when the standard error of the estimate is very small compared to the range of test values.

If the test value is determined "positive" with a value of 6 and the standard deviation/error is 0.5, then 6 works as a cutoff if a value in the range of 3-4 is thought to give a clear indication of doping use (and let us asume that the population average is 1.0 with a variation in the population of +/- 0.5 (0.5 is several (4?) standard deviations of a different distribution parameter than the standard error and is why I used 3-4 rather than 4 or three because we have a variation in the true level that reflects doping). If you increase the precision to 0.05, you can probably cut your critical value for the test down to 4.0 or just above. With this smaller SD you will have cut down substantially on the false negatives by moving the standard while still doing better on false positives. I suspect that t/e ratios are like this but there is the complication that both t and e are involved with "random terms"; the values used in the example are also genreally consistent with the testosterone test.

EPO has a relatively "big" "standard error" in the sense discussed above; in addition, it is not a 'univariate' test but one with blotches and other messy things. There are lots of places where errors can be made, since there are a lot of steps. Many (but not necessarily most) errors would likely lead to greater difficulty in detecting EPO through the test (this is the argument made by some in the supposed LA tests of the "B" samples for EPO when they had not been tested for EPO before and had been frozen for five or six years).

Sorry this is not better constructed and/or thought out/corrected, but I am hurried at work.