DBrower, idiot at large

Chances of a false positive are high when the incidence of real positives is low. But when the real incidence of positives is high (ie most are doping), chances of a false positive are low.

Notwithstanding that cyclists are masking/cooking their biology so that they don't test positive.

I don't understand his logic that most are doping, but those that are caught are unlucky false positives. I guess he is saying that everyone is able to fudge pass the test, so those that are caught are just the unlucky false positives? It is an interesting concept, but it relies on a lot of assumptions regarding the accuracy of the tests.

Crankyfeet said:

Chances of a false positive are high when the incidence of real positives is low. But when the real incidence of positives is high (ie most are doping), chances of a false positive are low.

Notwithstanding that cyclists are masking/cooking their biology so that they don't test positive.

I don't understand his logic that most are doping, but those that are caught are unlucky false positives. I guess he is saying that everyone is able to fudge pass the test, so those that are caught are just the unlucky false positives? It is an interesting concept, but it relies on a lot of assumptions regarding the accuracy of the tests.

Landis must be having a good ol fashion chuckle at his expense.

1. Alpha pro sport. If most dope, there comes a tipping point, where it becomes virtually all. Pros dont lose if they have an option. Thus, they justify their culture, and their own internal rules.

2. Doping only exists because 1) it works, and 2) it can be masked and rules subverted.

Folks like Tom Fine and Dave Brower, Michael Akinde (Strategy) look for the most specious of supporting evidence. The simple evidence is the power numbers on climbs, and the scientific studies. A good doping program gives you 15% on that finishing climb. It may be less on a chrono.

Everyone appreciates these athletes are off the curve, there is less then 1% seperating them in pure physiological terms, they all train hard, they all watch their nutrition and diet, they all have sports science advisers.

Those numbers are stark. Then you look at the entire GC winners and podiums in the last decade. Those with actual positives or actual links to doping doctors or doping rings, is near 100%.

So, if you are clean, giving up the 15% power to a doper on a final climb, just how can you get anywear near the podium. I think the last clean guy to ride into the top 10 of a GT was probably Mcgee in either 2004 or 05 at the Giro.

I still don't think you can win a doyenne clean. I dont think you can win the worlds tt clean neither. You may be able to get on the podium. Wiggins and Mcgee could almost snag a podium in the worlds tt.

thunder said:

So, if you are clean, giving up the 15% power to a doper on a final climb, just how can you get anywear near the podium. I think the last clean guy to ride into the top 10 of a GT was probably Mcgee in either 2004 or 05 at the Giro.

I still don't think you can win a doyenne clean. I dont think you can win the worlds tt clean neither. You may be able to get on the podium. Wiggins and Mcgee could almost snag a podium in the worlds tt.

Uncle Chris Boardman wasn't that long ago.

classic1 said:

Uncle Chris Boardman wasn't that long ago.

Can you do it now? The worlds tt was pretty weak back then, in depth and in talent. The best GC riders were the best tters. Ullrich, Lemond, Indurain. But the worlds tt is only a decade old, before it was GP des Nations, unofficial. Jens Voigt used to have a monopoly on that.

But like Michael Rich, and Uwe Peschel, Jens is a great tter, but not the best in the peloton. Even Rogers, great tter, but he was behind Ullrich. And behind many when he had to tt within a GT.

Now no one can compete with Cancellara. Remember Cancellara won the jnr worlds twice, and in the second tt, was faster kmph than Ullrich who won the pros, albeit 50% longer tt.

thunder said:

Can you do it now? The worlds tt was pretty weak back then, in depth and in talent. The best GC riders were the best tters. Ullrich, Lemond, Indurain. But the worlds tt is only a decade old, before it was GP des Nations, unofficial. Jens Voigt used to have a monopoly on that.

I'd argue the fields in Boardmans time (1994 to early 2000's) had better fields. Names like Olano, Indurain, Rominger, Zulle, Gontchar, Mauri crop up. Look at the podiums. Voight only won GP des Nations once.

Boardman was up against some outstanding TT'ers in his era.

The hour record for example was swapped between Indurain and Rominger in the mid-1990's, when they both reached the record at Bourdeaux.

Boardman and Obree also had hour record in that era as well.
Add Ullrich, Peschel, Zulle, Olano in to the mix and you've got a pretty select group competing with/against Boardman.

thunder said:

Landis must be having a good ol fashion chuckle at his expense.

1. Alpha pro sport. If most dope, there comes a tipping point, where it becomes virtually all. Pros dont lose if they have an option. Thus, they justify their culture, and their own internal rules.

2. Doping only exists because 1) it works, and 2) it can be masked and rules subverted.

Folks like Tom Fine and Dave Brower, Michael Akinde (Strategy) look for the most specious of supporting evidence. The simple evidence is the power numbers on climbs, and the scientific studies. A good doping program gives you 15% on that finishing climb. It may be less on a chrono.

Everyone appreciates these athletes are off the curve, there is less then 1% seperating them in pure physiological terms, they all train hard, they all watch their nutrition and diet, they all have sports science advisers.

Those numbers are stark. Then you look at the entire GC winners and podiums in the last decade. Those with actual positives or actual links to doping doctors or doping rings, is near 100%.

So, if you are clean, giving up the 15% power to a doper on a final climb, just how can you get anywear near the podium. I think the last clean guy to ride into the top 10 of a GT was probably Mcgee in either 2004 or 05 at the Giro.

I still don't think you can win a doyenne clean. I dont think you can win the worlds tt clean neither. You may be able to get on the podium. Wiggins and Mcgee could almost snag a podium in the worlds tt.

Interesting you mix Strat to idot as DB…Strat (I think he rr alter bad side) loves Riis and not objective…hates Lemond………… but not apologist as cheap prostitute DB....

Crankyfeet said:

Chances of a false positive are high when the incidence of real positives is low. But when the real incidence of positives is high (ie most are doping), chances of a false positive are low.

Notwithstanding that cyclists are masking/cooking their biology so that they don't test positive.

I don't understand his logic that most are doping, but those that are caught are unlucky false positives. I guess he is saying that everyone is able to fudge pass the test, so those that are caught are just the unlucky false positives? It is an interesting concept, but it relies on a lot of assumptions regarding the accuracy of the tests.

I don't agree with you there. There is usually no simple correlation between incidence of positives and the probability of false positives. I'm not saying that they are completely independent. But you can have a situation of high incidence of positives along with high chance of false positives, and low incidence of positives along with low chance of false positives. It depends on the criteria used to define a positive in a doping test. If you use stringent criteria, I would think that it is possible to make the probability of false positives very low (whether or not you have a clean peloton).

classic1 said:

I'd argue the fields in Boardmans time (1994 to early 2000's) had better fields. Names like Olano, Indurain, Rominger, Zulle, Gontchar, Mauri crop up. Look at the podiums. Voight only won GP des Nations once.

Olano contested, did the others contest the worlds consistently?

TheDarkLord said:

I don't agree with you there. There is usually no simple correlation between incidence of positives and the probability of false positives.

I didn't understand that either. I would think the probability of the false positives is almost entirely due to the particulars of the test (i.e. how well does the test detect what it is suppose to, and what factors influence the possibility of false positives).

However from what we know it seems pretty unlikely there is an unacceptable level of false positives because the false positive rate is related to the amount of false negatives and we know there are whole bunch of false negatives. IOW, the "sensitivity" of most dope tests is quite low (intentially so to avoid false positives). This is why so many dopers can avoid testing positive, the tests are set up to have a low sensitivity and high specificity. If OTOH, guys who doped were testing positive all the time (which they aren't since only a very small percentage of tests are positive and yet we know up until quite recently lots and lots of riders were doping) you could be fairly sure you would also have a relatively high rate of false positives (i.e tests with high sensitivity but low specificity).

limerickman said:

Boardman was up against some outstanding TT'ers... Boardman and Obree also had hour record in that era as well.

Hey Lim, why was Obree so good at the hour record and pursuit, but in every other aspect of the sport he was something less than an average pro rider?

helmutRoole2 said:

Hey Lim, why was Obree so good at the hour record and pursuit, but in every other aspect of the sport he was something less than an average pro rider?

UK cycling tt culture + Armstrong specialisation phenom.

You know Pozzato was just as good as Cancellara and Rogers in the chronos in the espoirs?

So, point is, if he doped, and did ride road races, he probably had some potential.

Wiggins hasa never looked like winning a road stage, unlike Mcgee.

helmutRoole2 said:

Hey Lim, why was Obree so good at the hour record and pursuit, but in every other aspect of the sport he was something less than an average pro rider?

I agree with Thunder : British concentration on the hour record (ie Boardman/Obree).

Also Obree was offered a professional contract at Le Groupement - but he has said that he found life in a professional team difficult and left.
(a couple of posters here, who know Obree, say that Obree was not prepared to dope aand that's why he decided to abandon professional road racing).

Crankyfeet said:

Chances of a false positive are high when the incidence of real positives is low. But when the real incidence of positives is high (ie most are doping), chances of a false positive are low.

TheDarkLord said:

I don't agree with you there. There is usually no simple correlation between incidence of positives and the probability of false positives. I'm not saying that they are completely independent. But you can have a situation of high incidence of positives along with high chance of false positives, and low incidence of positives along with low chance of false positives. It depends on the criteria used to define a positive in a doping test. If you use stringent criteria, I would think that it is possible to make the probability of false positives very low (whether or not you have a clean peloton).

I can illustrate what I was saying better with math for you TDL, rather than English, which may have been ambiguous.

Let's assume that the drug tests are 99% accurate (there is a 1% chance of a false positive) and let's assume two scenarios.

=> Scenario 1: The peleton has a low real incidence of doping and only 0.5% of cyclists are doping.

=> Scenario 2: Vastly different and 80% of the peloton are doping. Let's assume that the doping corresponds to what's being tested.

If we wish to determine the posterior probability that a given positive is a false positive, in each case, we can apply Bayes Theorem.

Scenario 1. (low real incidence of doping - only 0.5% dope)

Chance that a positive is in fact a real positive = (0.99 x 0.005)/[(0.99 x 0.005) + (0.01 x 0.995)]

= 0.332

therefore only 33.2% of positive tests are actually dopers and there is a 66.8% chance (1 - 0.332) that a positive test is a false positive.

Scenario 2. (relatively high real incidence of doping - 80% dope)

Chance that a positive is in fact a real positive = (0.99 x 0.80)/[(0.99 x 0.8) + (0.01 x 0.20)

= 0.997

therefore 99.7% of positive tests are actually dopers and there is a 0.3% chance (1 - 0.997) that a positive test result is a false positive.

Crankyfeet said:

I can illustrate what I was saying better with math for you TDL, rather than English, which may have been ambiguous.

Let's assume that the drug tests are 99% accurate (there is a 1% chance of a false positive) and let's assume two scenarios.

=> Scenario 1: The peleton has a low real incidence of doping and only 0.5% of cyclists are doping.

=> Scenario 2: Vastly different and 80% of the peloton are doping. Let's assume that the doping corresponds to what's being tested.

If we wish to determine the posterior probability that a given positive is a false positive, in each case, we can apply Bayes Theorem.

Scenario 1. (low real incidence of doping - only 0.5% dope)

Chance that a positive is in fact a real positive = (0.99 x 0.005)/[(0.99 x 0.005) + (0.01 x 0.995)]

= 0.332

therefore only 33.2% of positive tests are actually dopers and there is a 66.8% chance (1 - 0.332) that a positive test is a false positive.

Scenario 2. (relatively high real incidence of doping - 80% dope)

Chance that a positive is in fact a real positive = (0.99 x 0.80)/[(0.99 x 0.8) + (0.01 x 0.20)

= 0.997

therefore 99.7% of positive tests are actually dopers and there is a 0.3% chance (1 - 0.997) that a positive test result is a false positive.

Wow, CF, you're smarter than I thought.:p You're exactly right, false positives are dependent on the prevalance ("real incidence" in australian) of the condition in your population and the sensitivity and specificity of your test. That's why a B sample is drawn which will significantly lower your rate of false positives. The other technique is to use a different confirmatory test such as the IRMS for testosterone. In Flandis' case, he had a more sensitive test conducted twice (ratio test) followed by a very specific test (IRMS). What you hope for in this type of diagnostic testing is a high positive predicitive value - that a positive test accurately reflects the condition - in this case doping.

The problem with dope testing is that they usually have a very low negative predictive value - in other words a negative test is not indicative of someone being clean.

Crankyfeet said:

I can illustrate what I was saying better with math for you TDL, rather than English, which may have been ambiguous.

Let's assume that the drug tests are 99% accurate (there is a 1% chance of a false positive) and let's assume two scenarios.

=> Scenario 1: The peleton has a low real incidence of doping and only 0.5% of cyclists are doping.

=> Scenario 2: Vastly different and 80% of the peloton are doping. Let's assume that the doping corresponds to what's being tested.

I don't agree with your analysis here, and it has to do with how you interpret "test accuracy" with conditional probabilities. Your analysis assumes that the definition of "accuracy of test" = Probability(+ve test/doping) where I'm using the standard conditional probability notation, i.e. probability that test gives a positive given that athlete is doping. If that were 99%, no wonder your false positive rate is so high because the thresholds for triggering off positives has to be very low to catch all the dopers. I would say that the real definition of test accuracy is P(doping/+ve test), i.e. probability that the athlete is doping given positive test result. In other words, if you say that a test is 99% accurate, then the probability of athlete being clean given a positive test is 1%; or false postive rate is 1%.

A completely separate parameter is P(+ve test/doping), which measures how efficient a test is in catching the cheaters. Now, from Bayes theorem, P(+ve test/doping) * P(doping) = P(doping/+ve test) * P(+ve test). The last quantity, P(+ve test) measures the incidence of positive tests. P(+ve test/doping) is dependent on the test itself, as is P(doping/+ve test). It is INDEPENDENT of the fraction of the peloton that cheats.

Any dope tester would be seriously concerned if your numbers were really true.

Brower is basking in the glow of "celebrity". He's also too proud to admit that he's wrong. This is not news, and it's the same song other apologists sing.

He's a cynic like I'm a Landis supporter. Much like CampyBob is a "realist".

TheDarkLord said:

I don't agree with your analysis here, and it has to do with how you interpret "test accuracy" with conditional probabilities. Your analysis assumes that the definition of "accuracy of test" = Probability(+ve test/doping) where I'm using the standard conditional probability notation, i.e. probability that test gives a positive given that athlete is doping. If that were 99%, no wonder your false positive rate is so high because the thresholds for triggering off positives has to be very low to catch all the dopers. I would say that the real definition of test accuracy is P(doping/+ve test), i.e. probability that the athlete is doping given positive test result. In other words, if you say that a test is 99% accurate, then the probability of athlete being clean given a positive test is 1%; or false postive rate is 1%.

A completely separate parameter is P(+ve test/doping), which measures how efficient a test is in catching the cheaters. Now, from Bayes theorem, P(+ve test/doping) * P(doping) = P(doping/+ve test) * P(+ve test). The last quantity, P(+ve test) measures the incidence of positive tests. P(+ve test/doping) is dependent on the test itself, as is P(doping/+ve test). It is INDEPENDENT of the fraction of the peloton that cheats.

Any dope tester would be seriously concerned if your numbers were really true.

Au contraire.

P(doping/+ve test) is dependent on P(doper).

That was my point in the example I gave.

TheDarkLord said:

I don't agree with your analysis here, and it has to do with how you interpret "test accuracy" with conditional probabilities. Your analysis assumes that the definition of "accuracy of test" = Probability(+ve test/doping) where I'm using the standard conditional probability notation, i.e. probability that test gives a positive given that athlete is doping. If that were 99%, no wonder your false positive rate is so high because the thresholds for triggering off positives has to be very low to catch all the dopers. I would say that the real definition of test accuracy is P(doping/+ve test), i.e. probability that the athlete is doping given positive test result. In other words, if you say that a test is 99% accurate, then the probability of athlete being clean given a positive test is 1%; or false postive rate is 1%.

A completely separate parameter is P(+ve test/doping), which measures how efficient a test is in catching the cheaters. Now, from Bayes theorem, P(+ve test/doping) * P(doping) = P(doping/+ve test) * P(+ve test). The last quantity, P(+ve test) measures the incidence of positive tests. P(+ve test/doping) is dependent on the test itself, as is P(doping/+ve test). It is INDEPENDENT of the fraction of the peloton that cheats.

Any dope tester would be seriously concerned if your numbers were really true.

Cranky is right. Here is a brief explanation of positive predictive value. en.wikipedia.orgPositive predictive valueOpen ↗