Saturday, August 16, 2008

Saugy's Page 26

In response to our post about Ashendon's "highly, highly unlikely" comment, we offered a figure from a presentation from Saugy of the Swiss WADA lab that challenged Ashendon's stated assumptions.

Figure 1: Saugy's page 26.


In response, there have been some questions where we obtained the numbers, and the answer was first by eyeball (proven good enough for Brenna), and now, by counting.


Figure 2: Count the dots, 1, 2, 3...

From this, we get 32/156= 20%, which sort of validates eyeballing as a methology (± 20%), at least in this case dealing with discrete samples.

Then, there is the suggestion that we should not look at the overall, but the specific case offered by Landis, of starting around 15.5 and ending up around 16.1.

Figure 3: How many that started around 15.5 ended up around 16.1?

From this, we'll suggest that 6/35 = 17%, which isn't quite 20%, but is certainly well above zero. One rejoinder to this is that we've drawn the boxes too wide, and that we should look only at individual columns and rows. We believe that would be an exercise in futility given the low number of data points and is overconfident in the margins of error in the source data.

At what point is something "highly, highly unlikely?" TBV professionally does highly-available computer systems, working between four and five nines as target reliabilities -- 99.99% and 99.999%. A motto here is "With a one gigahertz processor, one in a million is a thousand times a second." Whole percentages are not, to us, "unlikely", but virtual certainties. It's a matter of perspective.

Given that one of the reasons guys become GC contenders is because they recover well, one could argue the odds of a highly placed rider being in that group are higher too. Just at raw numbers, we've got 32 riders above the line, which is more than one per team. Who on each team is it mostly likely to be?

But...

Another argument has been made that, essentially, the 25, er, 20% who went up were doping. While that is an improvement over the sentiment "they are all doping", it is unprovable. The purpose of Saugy's presentation was to identify things that needed to be considered in implementation of a "passport" program, and he didn't suggest these were anything beyond data to be considered -- he made no accusations.

Finally, there has been no counter-argument to the observation that the first week of a tour is a "rest" week for riders who did intensive training the week before. In a rest week after hard training, it is a reasonable natural reaction to have these values go up. We suggested and maintain that Ashendon's assumption that the first week of the tour is a "hard week of cycling" is physiologically incorrect, and therefore invalidates his conclusion. It does not disprove his conclusion -- there are still reasons to look further, which we acknowledge and accept.

We remain convinced that
  • The comment "highly, highly unlikely" is hyperbole, because we have shown a substantial likelihood of natural occurance, and shown one of Ashendon's premises to be invalid;
  • Ashendon is correct to draw no conclusions from the available data, a point missed by the strident.
  • Ashendon is correct to suggest more study would have been needed.
  • Additional blood tests done on Landis during the Tour also provided no conclusions.
  • If the UCI didn't do even more blood testing, that is not Landis' responsibility.
  • No agency or party has officially or openly suggested Landis did any specific oxygen vector doping because doing so would be inappropriate and legally unsupportable.

This will not stop people from making insinuations, but this discussion is now available for those who wish to look further.

32 comments:

Larry said...

First, I agree. Second, I wish I'd said it this well. Third, and I know it's kind of "dancing monkey" of me to ask, but how do you produce these charts?

DBrower said...

Normally I use GIMP (because I'm too cheap to buy photoshop). If nothing else, doing things for TBV has forced me to learn a lot of PDF, Excel, and GIMP that I never knew before.

In this case, the dots were colored in while I counted using MSPAINT (start->run->mspaint) because it's bucket fill did the right thing and the one in GIMP didn't. Or I couldn't figure out how to use it.

TBV

m said...

Hubris.

Ashenden is a world expert on blood doping. Yet you all who apparently know nothing about statistics and little to nothing about blood science assert that Ashenden's statement is "hyperbole".

LINEAR REGRESSION TUTORIAL

Eyeballing Saugy's dataset one can see that on average HB decreases over a week of competition. So how unusual is the fact that Landis's HB values increases from a start value of 15.5 to 16.1 over a week in the Tour de France? To answer this question statistically one typically uses linear regression. One caveat to any conclusion is that we really don't know what Saugy's dataset represents and how it was selected. Could there be dopers in this dataset, what sort of cyclists were sampled, how intense was the race or activity monitored, and most importantly what was Saugy trying to show by this dataset?

Given a typical HB start value, linear regression estimates what the average HB should be after a week of competition for the sample data points in the Saugy dataset. The data is repesented by HB start values on the X axis (horizontal) and HB end values on the Y axis (vertical). Regression estimates a line drawn through the data that minimizes the vertical distance from each data point to that line. (Actually it minimizes the squared distance but one can think about it as simply the vertical distance). This gives the best estimate of what the ending HB value should be (dependent variable) for any particular starting HB value (independent variable). Using this regression line, one can estimate what the Landis predicted ending HB value should be for his start value, assuming he was an average competitive cyclist in the Saugy dataset. One then compares Landis' actual ending HB value to his predicted ending HB value. If it is 2 standard deviations away one can say that there is 5% chance in a 100 that his ending value was due to chance. That is it was 95% likely due to some systematic deviation, e.g. doping. A standard deviation is a measure of how much the data in general deviates from the regression line. Typically one can draw a parallel line on each side of the regression line that is two standard deviations away. This is the typical measure of statistical significance.

Without knowing the value of each data point one cannot estimate the regression line or the standard deviation of the estimate. In principal one could estimate the values for each data point in the dataset from the graph and then calculate the linear regression. That is too much work for me, so we are left to eyeball the data to try to find a line through the data that best measures its central tendency. We can tell that the regression line is going to be roughly parallel but below the 45 degree center line because we can see that on average the data points cluster below the 45 degree center line. That is on average HB decreases over a week of competition. We can also see (this is a little harder and requires some experience in interpeting scatter plots) that the slope of the regression line will be less than 45% because there are relatively more data points of HB values below the center line as the starting HB value increases over 16 HB on the X axis. This tilts the regression line a bit to be flatter than 45 degress, perhaps 40 degrees as a guess.

So if we draw a line of about 40 degrees through the 15.5, 16.1 position on the graph and count the number of data points above that line and divide by the total number of data points we get a rough estimate of how unusual the Landis value is when compared to the average. It may be less or more because we don't really know how dispersed the data is. When I do this I get an estimate of from 7% to 14% depending where I draw the line.

Not too far off the 5% statistical significance figure, and certainly not enough to claim that Ashenden's statement that the increase in Landis' blood values was highly unusual is "hyperbole", especially given that Ashenden must be assumed to know what he is talking about.

Tenerifed said...

Which race is Figure 1. from? 2004 Vuelta a Espana? 2005 Tour de France? Another race full of dopers? Races full of doping make Figure 1. irrelevant. That can't be argued away. I am not the one saying Figure 1. means something. It don't mean anything.

The first week of the Tour de France is not a "rest" week. It is the Tour de France fer chrissakes. It is intense and hard for everyone. Hard racing with an hour time trial too.

highwheel said...

@ tenerifed
You should know that the first week is used as training for the overall contenders. This should show you how hard the first week at the Tdf is: the overall riders can hide in the bunch, they do not have to work, they only have to make sure that they do not loose time to the other GC riders.

Generally
Discussing about the hemoglobin values. First of all with inventing the blood passport the UCI took care that a standardization of blood taking, transport and analyzis was developed by the participating Anti Doping Labs. This has a reason: Before the Blood Passport Programm the blood tests of the UCI were not an Anti Doping Tool and as such not of a very high scientific and reliable standard. Those tests were conducted with different machines, partly in field, partly sent to Kliniks (not WADA accredited) or Anti Doping Labs. Conditions of transportation varied, machines varied. You also should think about medical conditions, e.g. diarrhea- not uncommon in races or the use of antibiotics, more early and often prescribed in athletes than in average population because of the importance to stay in the race. And an unwelcome but frequent side effect is diarrhea...

It's not black and white. If it would be as easy as some of the people think you wouldn't need standardization and evaluation.

verifythentrust said...

It's amazing that is so easy to refute a world expert with a simple graph.

wschart said...

"Highly, highly unlikely" is such a vague statement it is almost impossible to refute. If I were to say that it is highly highly unlike for X to happen, and you were to show me that X actually occurs 10% of the time, so you disagree with my assessment, I can counter by saying I think anything under 15% is "HHU".

Ashendon may be an expert, but we don't know on what basis he makes this statement. Is it simply an opinion, or is it based on research that he or others have done?

Thomas A. Fine said...

Hubris?

Maybe.

I haven't complained when people question Berry's work. I point out the flaws in their arugment, but as for the hubris, I actually applaud anyone who questions "authority". Because they are just people. They're supposed to have the expertise to always
say the right thing, but nobody's perfect.

My involvement in all of this really got a head of steam when the head of the LNDD laboratory said of the CIR testosterone test:

"It's foolproof. This analysis tells the difference between endogenous and exogenous,"

This is a ridiculous statement, and verges on an outright lie. As the head of the lab, he should know better. Most of the world just trusted him because of his position.

So we are all entitled to all the hubris we can muster. Still, that hubris has to be backed up by something. Like actual data, which is a much more sound way to form an argument than the usual "trust common sense" approach (for an example tenerifed's belief that the first week of the tour is hard, even though available data proves that in relative terms it is not).

At any rate, two standard deviations is not at all a safe distance for a diagnostic test. 1 in 20 will fail through natural variations.

Also, regarding changing the angle, this may not be the right approach for this type of plot. The distance above the sloped line is the delta-Hb. To set a line with a standard deviation distance away you want to do that in absolute terms not relative terms. So the line should be raised.

Even if angle was the right approach, you'd be changing the angle at the true origin (0,0), not the displayed origin (12,12).

tom

Tenerifed said...

I present to you, Tyler Hamilton's 2003 in doping according to Danish newspaper Politiken: Well it started in 2002 actually where he 12 times took EPO, from the 21st of December to the 8th of January (conveniently enough having a break at the 25th and the 31st, even riders have holidays). The quantities vary through the calendar, it's either 1000 or 2000, those in the start of the year were 2000.
At the 14th of January he had his blood tapped for the first time, he then had a 10 day break before starting on a combined treatment of anabolic steroids and EPO, taking anabolic steroids every day from the 24th to the 31st, and taking EPO 3 times during period. From the 2nd of February to the 8th, he took EPO every 3rd day (quantity 2000).
A new period again started off by first tapping blood, then getting he old blood reinserted and then taking both EPO and anabolic steroids, all in the same day. He continued the treatment with EPO and anabolic steroids every day for 12 days, throwing in a blood analysis too.
To prepare for Paris - Nice he begun taking Gonadotropines (A term for hormones that among other things increase your testosterone production), and a little shot of Epocrin (Russian EPO). This treatment stopped two days before the start of P-N; he also had his blood reinserted at that date.
Hamilton won the mountain jersey in the race, but otherwise didn't make that much of a mark. The day after the race he had his blood tapped twice, a shot of growth hormone and some EPO.
The next period saw a treatment of EPO, growth hormone and insulin; he took it every other day up till the start of Criterium International, still taking growth hormone during the race. After the race e continued with growth hormones and insulin, also taking a day of taking and reinserting blood, just three days from the start of Pais Vasco.
The first day of the race was marked off with a shot of growth hormone, and also ended with a shot of growth hormone. He finished second to Iban Mayo, but most were of the perception that he didn’t go 100% in the final rain wet TT, and therefore lost the race there. The next days he continued with the growth hormone and got his first blood doping too. He rode Fleche Wallone, but finished only 14th. This started a period with no doping for 13 days, but with plenty of success on the road as he both won Liege – Bastogne – Liege and Tour de Romandie.
The day after his Tour de Romandie victory was celebrated with the ritual tapping of blood, and then the following days a massive EPO cure; from the 6th of May to the 22nd he took EPO and anabolic steroids almost every day. And then gradually slowing down his intake of EPO, till the 2nd of June where he got his blood tapped and reinserted, he subsequently got some growth hormone the 3rd and the 5th, exactly a month before the Tour.
He rode Dauphine Libere without taking anything, but finished the race with the usual tapping and reinserting of blood. The finish of the race also marked the start of the final preparations for the Tour. Setting off with growth hormone and then gradually taking some Epocrin and Insulin Growth Hormone(IGH). His last intake of IGH was at the 1st of July, a mere 4 days before the start of the Tour. At the 2nd of June he got his blood tapped twice and got 3 reinsertions of blood, he also got one on the 3rd, the same day he got tested by the Tour de France management. His two blood insertions during the Tour were at the 11th, the day before the first Alp stage and at the 17th the day before the first big TT. Hamilton won the 16th stage, the 23rd of July and finished 4th in the race, despite a broken collarbone.

Tenerifed said...

The Suddeutsche Zeitung has published what it claims was Jan Ullrich's doping plan for the first week of the Tour de France 2005, based on information from Operacion Puerto. For the first seven days of the Tour, Dr. Fuentes prepared him a series of hormones, insulin, cortisone, testosterone and blood transfusions, the newspaper alleged. It claimed that there was a "Roadbook" for the Tour 2005, which investigators link to Ullrich.

On the first day, according to the SZ, the hormone HZ was listed, the second day insulin I-3, the hormone TGN and cortisone, the third day TGN and PCH (a testosterone shot), on the fourth day HMG, a hormone mixture, a "rest day" on the fifth day, the sixth day insulin I-3, and on the last day, the re-infusion of his own blood, as well as insulin I-3 and vitamin E.

http://tinyurl.com/6duorz

Russ said...

Wow, great discussion.

M said:
"One caveat to any conclusion is that we really don't know what Saugy's dataset represents and how it was selected."

and

Highwheel said:

"Before the Blood Passport Programm the blood tests of the UCI were not an Anti Doping Tool and as such not of a very high scientific and reliable standard. Those tests were conducted with different machines, partly in field, partly sent to Kliniks (not WADA accredited) or Anti Doping Labs. Conditions of transportation varied, machines varied. You also should think about medical conditions, e.g. diarrhea-"

And don't forget to add the other variables TBV brought up earlier like variations from sample taken while standing, sitting, laying, dehydration variations from more that diarrhea sources - like sweating and limited availability of re hydration.

And I hope that I am correct in interpreting the Higwheel quote as placing Floyd's results squarely in the center of the caveat class!

So M, how is it that a rigorous regression analysis is worth even thinking about?

Regards,
Russ

Tenerifed said...

Russ-

Highwheel is wrong. Hematocrit limits have been an Anti Doping Tool since 1997. If you go over the limit you are banned for 2 weeks. The UCI knows what they are doing.

TBV-

Will you make a blog post that shows published doping plans for Tour De France contenders?

Russ said...

Tenerifed,

Ok, so lets try correcting the part of highwheel I quoted by removing the words:
"not an Anti Doping Tool and as such"

Now we are left with the rest and the fact that Floyd did not exceed the UCI limit. So we are in the data noise, so to speek. Now we are attempting to squash this popular notion that Floyd is guilty of EPO or the like on the basis of statics which I am sure the all knowing UCI took into account in establishing the criteria that accepted Floyd's results and allowed him to continue.

So the question still stands, how is it worth the effort of the regression analysis when Flyod's data can be excluded from the eyeball'd 95% regression quilty that M came up with? esp since his data has mitigating questions that place him in the caveat catagory.

Regards,
Russ

wschart said...

So, where exactly does this alleged doping plan of Hamilton's come from? I doubt from him, AFAIK he is still denying his guilt.

Tenerifed said...

Russ-

A regression analysis is worthless. The chart means nothing to this argument because many cyclists on the chart were doping.

Read the stories of Tyler Hamilton and Jan Ullrich's doping plans. Blood doping has been business as usual. Look at the stuff Ullrich was going to do in the first week. He was going to blood dope and take testosterone too. Highly usual.

Wschart-

Tyler Hamilton's home address in Colorado was written on the back of his plan. The bill for it was faxed to his wife.

DBrower said...

The UCI tests are not "anti-doping"; they are officially a "health-check". The two-week no-start is not an anti-doping penalty, it is "for the riders health"

As such, highwheel is right to note the health check tests are done with less rigor than official anti-doping tests.

TBV

Tenerifed said...

TBV-

They are a health check because EPO was killing cyclists. They should be called survival checks. Whatever you call them they are anti doping. They are done to reduce EPO use.

It was an anti doping penalty for Marco Pantani. He was in the Maglia Rosa when he got banned from the 1999 Giro.

DBrower said...

T-fed,

There is no question the UCI tests are/were a proxy for true anti-doping tests. At the same time, the sampling/measurement methods have not been setup to even the rigor of anti-doping tests, which was Highwheel's point -- they are, at best, at the "screening" level of accuracy. To call them "anti-doping tests" is to confuse them with tests taken under WADA protocols and sent to WADA labs, under WADA standards, which they are not.

Highwheel is right -- the UCI data is not reliable to an "anti-doping" standard, though it is fine as an indicator of the need to do followup, in the form of a no-start or, as seems the case, with targeted testing. Since nothing came of the other blood taken from Landis, it is not clear there is anything conclude.

TBV


TBV

Larry said...

M –

Now that I’ve decided that I don’t need to argue with you, and that I can work with you in an effort to explore matters of interest … well, I really enjoy your posts more than I used to.

I’ll let TAF carry the argument about on the statistics. I’ll only state that there must be something wrong with your statement that “there is 5% chance in a 100 that [Landis’] ending value was due to chance. That is it was 95% likely due to some systematic deviation, e.g. doping.” This must be like saying that if I score in the top 5% of my law school class (hah! As if!), that there was a 5% possibility that this would have been due to chance and a 95% chance that it was due to some systematic deviation, e.g., cheating. I just might be smart (hah!) or maybe I studied really hard (as if!). SOMEONE is going to be in the top 5%, that’s the nature of things, that’s not something that requires an explanation of some “deviation”. But … I’ll leave it to TAF to explain this in a better way than I can.

I’ll focus on whether we can question the experts. One obvious point: when the experts disagree, we have no choice but to question them if we’re trying to understand what’s going on. We don’t find many areas here where the experts agree. (Hey, I let YOU question Dr. Berry, and that was in a prestigious and peer-reviewed science journal!)

Also, I’m not questioning Ashenden’s world-class expertise. Notice that both TBV and I have paid a good deal of attention to Ashenden, in particular to his emphasis on the importance of the Offscore. Ashenden either invented the Offscore or helped invent it (not clear), and I have said that use of the Offscore is acknowledged as the best (but not the only) way to interpret blood readings. That shows respect for Ashenden and his expertise. I’m not trying to dismiss what Ashenden said out of hand, and of course I don’t think I know more than he does. I’m just trying interpret what Ashenden said in a reasonable way, consistent with the evidence, his past statement and statements made by other experts.

Let’s focus on the question of hyperbole. I believe that experts engage in hyperbole, just like the rest of us. I do not believe that everything an expert says is intended for scientific peer review. Sometimes, experts shoot from the hip.

Here’s another example of hyperbole. An expert made the following statement:

“Landis is just the latest in a string of Tour champions to be implicated in doping cases. He has a tangible opportunity to salvage his sport. He has a unique platform from which he can reveal what cycling has become, and in so doing compel his sport to face its demons and begin its journey down the road of self-healing.”

That’s a pretty strong statement that Landis is guilty. We might understand it if this statement was made recently, after the two hearings, or after the expert in question had examined the LNDD document package. But that’s not what happened. The expert in question is Ashenden, and his statement was reported on August 6, 2006. See 2006 San Diego Union Tribune Article. That’s a little early to be making pronouncements of guilt, don’t you think? Would it have been “hubris” to question Ashenden’s expert opinion on this point, to say that we might still want to review the evidence, or that Landis might still want to contest the AAF?

Experts are human and they DO engage in hyperbole, and sometimes they cross a line. As TAF pointed out, De Ceaurriz said that the CIR test was “foolproof” and that “no error is possible.” Now, THAT’S hyperbole! Of course a CIR error is possible. We won’t even mention the fact that virgins within the vicinity of the 2006 Tour were perfectly safe, at least from the riders. We might ask some of these experts to tone down their hyperbole, and we don’t have to take everything an expert says at face value.

I think that we would do a disservice to Ashenden if we plucked a particular comment (like the “highly highly” unlikely comment) out of the context of his entire quote. He also said in his quote that he could not say that Landis definitely doped and that he should have been targeted for further testing. I don’t have a problem with either of those statements.

In fairness, we should look at the context of the Ashenden statement. He did not make this statement before a court, or arbitration panel, or in a scientific journal. He made this statement in a telephone interview. See 2007 San Diego Union Tribune Article. I know he’s an expert, but an expert can make an off-the-cuff remark.

I think it’s fair to interpret Ashenden’s statements in light of his other published statements on blood tests. As far as I can tell (based on a google search, not infallible), the telephone interview is the only time that Ashenden has ever gone on the record regarding the Landis blood values, or about the unlikely nature of a Hg increase during the course of a Tour. Ashenden is (or was) the Project Coordinator of “Science and Industry Against Blood Doping, which is reportedly an international research consortium. He’s certainly had numerous opportunities to repeat and clarify his statements about Landis or urge that appropriate action be taken against cyclists who show Hg increases during the course of a Tour. As best as I can tell, he has never done so.

In Ashenden’s other published statements, such as in the recent Bicycling.com article, Ashenden de-emphasizes the importance of hematocrit readings (which would normally track Hg readings), in favor of the Offscore and reticulocyte readings.

Actually, the Bicycle.com article is a good example of why we need to be careful in considering the overall context of an expert’s statements. In this article, he said that some of the riders’ reticulocyte variations seemed abnormal. Would it be right to draw conclusions from this statement, out of context? Of course not. He also questioned the ACE data, which he was able to do because there was a second set of data from UCI. He also said that “there is nothing striking in the data as shown that indicates blood manipulation.”

I think it’s fair to read the Ashenden comments in light of the comments of other experts. For example, in the same bicycle.com article, ACE CEO Dr. Paul Strauss stated that “whenever we have a suspicious result, we also look at transport time, temperature and other aspects. We have to take that into account and not automatically raise all kinds of alarms if we get a suspicious result. There are other, technical reasons why results could be abnormal.”

If you read the Ashenden comments together with the Strauss comments, it would be fair to say that the Landis readings would justify further testing, but not the “raising of alarms” and certainly nothing like a reasonable suspicion of guilt. I think this gives due respect to Ashenden’s expert statement, to Strauss’ expert statement and to everything else we know.

Tenerifed said...
This comment has been removed by the author.
m said...

So what's the consensus?

Is the Saugy dataset garbage? If so, its very weak evidence to be used either to discredit or support Ashenden.

Tenerifed said...

TBV -

The hemoglobin and hematocrit of Floyd Landis are incriminating. He is banned for using testosterone. All these other doping plans used everything with blood doping. You can rationalize and parse more than Bill Clinton himself but it will not change the big picture.

What did you think when Emanuele Sella was rocketing around Italy? Did you think he was clean? Was he different from Floyd Landis? He is now because he had the guts to confess. Will you rationalize that Landis is a champion and Sella is pack fill? Sella needed drugs to rocket but Landis had talent and hard work. The truth is cyclists dope and more dope works. You take too much and it is obvious. Sometimes you get busted.

Larry said...

M –

Remember that Saugy is the head of the Lausanne WADA lab, with multiple peer-reviewed publications on doping to his credit. He is ALSO an expert, deserving of the same respect as Ashenden, and subject to question just like any other expert. But it’s a bit strange for you to say both that it’s “hubris” to question Ashenden and that Saugy’s data might be “garbage”. Especially when Saugy’s data is based on 154 data points and Ashenden’s data only two.

The web site containing the Saugy data appears to be down. I have a copy of the presentation containing this data. It is entitled “Some Principles For the Hematological Passport”, and the authors are Martial Saugy, Neil Robinson and Pierre-Edouard Sottas. The three of them have published together before, and I think they all work at the Lausanne lab. The presentation strongly advocates the use of the biological passport.

There is a section entitled “Yes BUT !!! Some Precautions have to be taken !!!” The first precaution is that “A unique blood protocol has to be Accepted and validated by WADA”. Saugy seems also to be recommending that a protocols be established for when and how the riders are tested, as there follows (1) data from a 1988 study by Leppanen and Grasbeck showing different Hct readings depending on whether the subject is supine, sitting or standing, and (2) a study by Robinson, Saugy et. al showing Hg and Hct increases before and after exercise. The chart displayed by TBV follows next. There is then a discussion of the effects of hypoxia on endogenous EPO production. There are then a couple of slides titled “Necessity to know the temperature of the samples over time”, apparently emphasizing the importance of proper transport and storage. Then there are two slides titled “Lack of standardization, reticulocyte count in particular”, comparing differences in various ways to measure reticulocyte counts (Abbott, Bayer, Sysmex). After this there are slides on how to take the right precautions (such as internal and external quality controls).

There is no explicit explanation for the inclusion of the slide reported by TBV in the Saugy presentation. Clearly, Saugy intended it to be evidence for the need for precautions. One obvious interpretation is that Saugy was trying to warn his audience that there is an observable rider-to-rider difference in plasma volume changes, even among the riders one would presume to be clean, as no two riders seem to measure out the same way. Possibly he was making some other kind of point that was supposed to be taken as a precaution.

I assume that you’ve attended as many Powerpoint presentations as I have. It’s not always easy to determine why a speaker includes a particular slide.

m said...

Larry,

TBV was using the Saugy dataset to factually question Ashenden's characterization of the change in Landis' HB values. I was trying to point out how a proper statistical analysis would be applied. I'm really only interested in talking about the statistics since many of you don't seem to understand this very well at all. Not that I'm an expert, but at least I've had some statistical training.

As to the Saugy data, it sounds like none of you know much about it. I know zero about it, and was trying to learn more. Clearly any inferences we can draw from it are limited by both the statistics and the nature of the data itself.

wschart said...

Saugy's data cannot be used to either confirm or deny that Landis blood doped. Ashendon himself says that the Landis data cannot be used to establish a doping violation, but merely as a basis for further testing (now a very moot point). Has there been an expert who claims the Landis figure are conclusive for blood doping?

Ali said...

tenerifed,

Construct one logical arguement, citing sources, which links together all of the points which you wish to make and explain how they apply to Floyd Landis. I'm genuinely interested in your views but until they are clarrified ... Sorry, but I work with people who are habitual bullshitters ... lay it down nice and simple and I may listen. Otherwise, I'm just hearing lots of jamming noise.

Ali

Larry said...

ws, and remember, Landis WAS tested further. 8 times total, I believe.

t-fed, props to you, buddy. You succeeded in bringing Ali back home where he belongs.

M, the rest of this is addressed to you.

True enough, all I know about the Saugy data is that Saugy is an expert, and that the data was included by Saugy as part of an overall recent presentation supporting adoption of the biological passport. I don’t imagine that Saugy would include “garbage” data in one of his presentations, at least not intentionally. (I am aware of how critical I’ve been of Saugy in the past.) At the same time, I doubt he was trying to show the exact Hg increase that would be considered to be usual or unusual. Possibly all he was trying to say is that there is some natural variation in blood marker movements among cyclists in the peloton. Still, this is the only data we have, and while I’m willing to take it with a grain of salt, I’m not willing to toss it out in favor of whatever similar but undisclosed data Ashenden might have had at hand during his telephone interview.

I’m arguing here that we have to be careful about drawing inferences from ANY of this kind of data. This is why I got involved in this discussion, not to argue about whose data is best, but to argue about being careful in reaching any conclusions from this data. In this regard, I like the statement I quoted earlier from Dr. Strauss about raising alarms over a single suspicious result, and I like reading the Ashenden and Strauss statements together. To be sure, Ashenden was very careful (and I think, responsible) in the statements he made at bicycle.com, and I think my cautious point of view is consistent with those statements of Ashenden.

This is my main point, caution in the interpretation of this data. It is a SECONDARY argument to say that the reported 0.6 increase in Hg does not appear to be unusual based on the Saugy data. In making this argument, I’m not trying to discredit Ashenden. I’m trying to understand what he meant by “highly, highly unusual” in a larger context.

OK, I can accept the criticism that the method I used to determine what is and what is not unusual may not be state of the art! But it is certainly relevant to say that 20% of the riders shown on the Saugy chart had measured Hg increases, and that 12% or 17% of riders like Landis showed Hg increases like Landis. Those are not false statements, and I’m not willing to toss these statistics out the window just because there may be a better way to compute the statistics that no one has been willing to do.

I understand that linear regression may be the better way to analyze the data. Actually, I think I understand linear regression in gross and general terms, it’s sort of a standard deviation analysis based on a two-dimensional determination of the mean (the trend line). So, instead of doing what I did (looking at the data above and below the diagonal line represented by a 0.6 increase in Hg), you’d instead look at the spread of results above and below a relevant point in the trend line (about where x= 15.5). If I understand it correctly, then what TBV did in figure 3 is a rough form of regression analysis, as the pink box in the figure focuses us on the data points near where x = 15.5. Yeah, there’s no way I’ll figure out where the trend line passes through the pink box, or guess where the 2 standard deviation line would be above that trend line, but if the two standard deviation line came out below the Landis data point, that would leave 17% of the data near x=15.5 above that 2 standard deviation line, and I don’t think THAT is usual.

There’s also the usual caveats about margin of error, etc., etc.

I’ll leave any further discussion of statistics to TAF. From me, it’s best to simply take away my earlier point about being careful not to rush to judgment with this kind of data, particularly when it’s based on two data points and one purportedly suspicious measurement.

Russ said...

Garbage database?

My take has been simply that any data collected for research and esp for setting or calling for standards, would be expected to have been collected in a controlled manor to minimize the error sources and shrink error bands. This beyond what would be done in the field for normal use.
See Larry's papers...

I too have not seen the Saugy db but used the above assumption in my reasoning. If that does apply to the database, then the criteria that is drawn from it would be tighter than a reasonable criteria that would be applied (by uci) to the blood values collected pre blood passport program. This being because of the lack of controls and care and etc etc.

So if this reasoning and the basis for it is correct then it would mean that Floyd's results would fit into the caveat class.

Now if the data were included simply to illustrate the authors point, and not to prove it, then it may in fact be practically garbage.

so caveat not empty

Ali, great to hear from you!!!

Russ

Tenerifed said...

Ali-

Very simple. There are 3 possibilities-

1. Landis did not dope.

2. Landis only used testosterone.

3. Landis had a common doping plan. He used steroids and other hormones and blood doping.


What do people believe about these probabilities?-

1. Many Trust But Verify posters and other Landis fans believe this.

2. Does anybody believe this? I doubt it.

3. Easy for cycling fans to believe. Read the doping plans of Hamilton and Ullrich I posted. They used boatloads of stuff. There are many more plans like those. News comes out that cyclist banned for test had a high hematocrit at the TDF. Easy to believe he was using a typical plan.

wschart said...

What Hamilton or Ulrich might have done has not probative value as to what Landis might have done. Yes, I know there's the argument that such a large part of the peleton doped that it would have been impossible for a clean rider to do well. But that depends on the relative strengths of riders without doping. If doping makes one 25% faster, then if I'm 26% better than you are clean, I could still beat you if you doped. In addition, there are factors like tactics, team support, etc. And if so many riders in 2006 doped that the only way Landis could have come out on top was to dope, then what about Periero and others? Why does no one suggest they must have doped when if you follow the logic that Landis must have doped to get his results, you must conclude that others did too.

Landis might have blood doped, no test can ever prove he didn't. But the blood values we have for Landis do not prove that he did.

Larry said...

tenerifed, I'll say it a bit differently than ws, but I think I mean the same thing:

I can't endorse a method of analysis that divides possibilities into three categories, usually consisting of two relatively distinct categories (one naive, one improbable), leaving a third category of sweeping generality as the seemiongly reasonable choice.

In other words, I can't go along with a line of reasoning that says that you doubt he didn't dope, and no one in his right mind would dope only with testosterone (an argument that USADA rejected when it was made by Landis proponents in the slightly different form of "why would a cyclist want to dope in-race with testosterone?"), leaving the only reasonable possibility that Landis must have doped using every technique imaginable.

I find your line of reasoning roughly parallel to dealing with a kid who you think stole the morning paper off your front lawn. You doubt that the kid could be innocent, and no one would be dumb enough to steal only the morning paper when you can just go to the local library and read it there, so the kid is probably guilty of every crime imaginable.

I wasn't even going to call you on this, because to be honest, you're just saying what I think 80% of the world is thinking. So I hope I don't sound too critical. I appreciate your honesty even if I think your reasoning is faulty.

Instead of your three-pronged approach, I prefer my own three-pronged approach:

1. Landis is guilty of whatever it is that they can prove he did.

2. He is innocent of the rest.

3. God bless the United States of America.

Ali said...

tenerifed, wschart, Larry,

Wschart's and Larry's three points pretty much summarise what we could all reasonably assume without prejudice (and admittably coincide with my own thoughts - although substitute USA with "Chris Hoy" ... I'm feeling nationalistic today).

Unfortunately, prejudice is never far from the human mind. I don't have anything to add really, as usual. I just don't see your arguement. As others have pointed out it's just "well, if others (allegedly) cheated, he must of, based on the fact that he DID NOT fail inconclusive, "quality control", snap-shot measurements which are known to have poor accuracy".

That sort of arguement doesn't have me measuring out the rope and tying the knot. Still, you're convinced, so good luck pursuing this ... ?

Ali