a shill by any other name wrote:
Word of caution about interpreting statistics (which you've done a very nice job of compiling and explaining)... it's easy to see patterns that aren't real, and assume that the strength of your dataset is adequate to defend your interpretation.
Thanks. Except for my choice of cutoff, to select "top time", I used a rather complete dataset.
We could do similar analysis, using top-10 average, top-20 average, or even top-20th time.
All the numbers would be bigger, and the percentages would change.
I'm not sure it would change the general observations though -- there would still be enough basis to ask, for 5 continents, why so few by so little when EPO promises so much?
a shill by any other name wrote:
rekrunner,
thanks for the interesting analysis. Word of caution about interpreting statistics (which you've done a very nice job of compiling and explaining)... it's easy to see patterns that aren't real, and assume that the strength of your dataset is adequate to defend your interpretation.
Interesting story (if perhaps only to me), a colleague recently shared with me a neat paper about linkages between lunar cycles and large earthquakes, which drew the (evidently well-defended) conclusion that very large (M8+) earthquakes show no relationship with lunar cycles (which correspond to small cyclical changes in gravitational pull, thus small changes in crustal stresses...). The paper got me wild-eyed interested in the topic so I did my own exploration, and grabbed about 14000 "almost random" earthquakes (M2.5+) from available records (USGS) to see if I could detect an underlying lunar signal. Those data DID suggest a relatively higher frequency of earthquakes around the new moon relative to the full moon (with a lot of scatter), so I explored further, grabbing over 50,000 earthquakes. This larger dataset seemed to confirm and strengthen the initial hypothesis, so I got really excited and shared the finding with some colleagues. One smart colleague asked me if the interpreted pattern was statistically significant, a fair question; so, working a little more methodically to choose more earthquakes randomly, I increased the dataset to 104,000 earthquakes. At this point, the evident pattern completely disappeared.
The lesson that I (re)learned in this exercise is that it's easy to fall in love with conclusions you've drawn from statistical datasets, but the reality of (particularly complex, chaotic) natural phenomena is that they like to defy probabilistic description in sometimes unexpected ways. And interpreting the presence or absence of a doping effect from a small number of performances is a risky thing to do. Or at least one that ought to be done without a high degree of confidence expressed in the outcome.
Nice work though. Cheers
Maybe so, but "complete" isn't remotely the same as statistically significant. In my counter example it took > 100,000 data points to disprove a false interpretation derived from > 50,000 data points. Here you are using a few dozen data points and presuming (because they capture "all" the top performances) that they represent the truth of all humanity. Big stretch... VERY big...
rekrunner wrote:... I used a rather complete dataset.
a shill by any other name wrote:
(I'm surprised you don't seem to "get" this, given the evident understanding conveyed in your first summary)
casual obsever wrote:
rekrunner wrote:
Splitting the improvements in smaller sub-groups highlights the regional nature of large improvements
It also worsens the statistics. A lot. And potentially skews them when the person who wants to make a point chooses the sub-groups, like you did here.
Also, you need to be more consequent. This here is a good start:
rekrunner wrote:
If these claims are accurate, we should be able to observe the large effect from EPO on the best performances, pre-1990, versus post-1990.
But if you consider that the best performances post-1990 are EPO-aided, then you need to consider that the pre-1990 performances are also enhanced by the then basically undetectable (legal and illegal) methods.
Otherwise you'd postulate that the sport was clean pre-1990. We know better.
The question then becomes: how much - possibly gender- and distance-specific - helps (at first unlimited) EPO and controlled but more powerful steroids vs. unlimited blood transfusions and basically unlimited but primitive steroids. And yet, each time somebody brings this up, you argue against it.
On the other hand, if you were to think that runners are too honest to cheat even without noticeable risk of detection (and not even use blood transfusions while allowed), then this method were absolutely pointless. Then we can only look at the athletes with positive tests, and whose doping history was proven by admissions and witnesses. At the top, those are 0, as nobody talks, and everyone pretends they only doped once.
Finally, let's look at the numbers above. Why would the longer distances benefit more? Ok, the marathon was underdeveloped in the 80s, and now there is more money, so we can't really compare that.
But do 5000 m runners earn more than 800 m runners? Doubt it. More likely, the benefit from the steroids decreases with increasing distance, while blood doping methods become more important.
C'mon man...they're not "my" EPO champions & I never said they were as fast as the pre-EPO era times. But there's a problem when you examine the fastest women's 800 times and conclude that's it's only androgens involved & not augmented by blood doping. 5 out of the 6 fastest times are pre-1990 but 3 out of those 5 and 6 out of the top 10 all-time best are runners from Soviet block countries. In these cases, you can't rule out blood doping as it was being used extensively with the Soviet block nations in the 70s & 80s. So, was it androgens only pre-1990? Blood doping & androgens? OTOH, could some of those non-Soviet block runners possibly be clean?
rekrunner wrote:
Your EPO champions were not as fast as the steroids champions of the past.
Good point on blood doping methods:
casual obsever wrote:
rekrunner wrote:
Splitting the improvements in smaller sub-groups highlights the regional nature of large improvements
It also worsens the statistics. A lot. And potentially skews them when the person who wants to make a point chooses the sub-groups, like you did here.
Also, you need to be more consequent. This here is a good start:
rekrunner wrote:
If these claims are accurate, we should be able to observe the large effect from EPO on the best performances, pre-1990, versus post-1990.
But if you consider that the best performances post-1990 are EPO-aided, then you need to consider that the pre-1990 performances are also enhanced by the then basically undetectable (legal and illegal) methods.
Otherwise you'd postulate that the sport was clean pre-1990. We know better.
The question then becomes: how much - possibly gender- and distance-specific - helps (at first unlimited) EPO and controlled but more powerful steroids vs. unlimited blood transfusions and basically unlimited but primitive steroids. And yet, each time somebody brings this up, you argue against it.
On the other hand, if you were to think that runners are too honest to cheat even without noticeable risk of detection (and not even use blood transfusions while allowed), then this method were absolutely pointless. Then we can only look at the athletes with positive tests, and whose doping history was proven by admissions and witnesses. At the top, those are 0, as nobody talks, and everyone pretends they only doped once.
Finally, let's look at the numbers above. Why would the longer distances benefit more? Ok, the marathon was underdeveloped in the 80s, and now there is more money, so we can't really compare that.
But do 5000 m runners earn more than 800 m runners? Doubt it. More likely, the benefit from the steroids decreases with increasing distance, while blood doping methods become more important.
a shill by any other name wrote:
rekrunner wrote:... I used a rather complete dataset.
Maybe so, but "complete" isn't remotely the same as statistically significant. In my counter example it took > 100,000 data points to disprove a false interpretation derived from > 50,000 data points. Here you are using a few dozen data points and presuming (because they capture "all" the top performances) that they represent the truth of all humanity. Big stretch... VERY big...
I guess there is no problem then, as I did not exclude transfusions:
"I'm not seeing a strong case for replacing transfusions with EPO boosting VO2max to new levels of performances, for men or women."
Rocket Fuel Madness wrote:
rekrunner wrote:
Your EPO champions were not as fast as the steroids champions of the past.
C'mon man...they're not "my" EPO champions & I never said they were as fast as the pre-EPO era times. But there's a problem when you examine the fastest women's 800 times and conclude that's it's only androgens involved & not augmented by blood doping. 5 out of the 6 fastest times are pre-1990 but 3 out of those 5 and 6 out of the top 10 all-time best are runners from Soviet block countries. In these cases, you can't rule out blood doping as it was being used extensively with the Soviet block nations in the 70s & 80s. So, was it androgens only pre-1990? Blood doping & androgens? OTOH, could some of those non-Soviet block runners possibly be clean?
Are Gary and Casual Observer actually different people? I somewhat doubt it, but if you are, couldn't you just have this mutal two way jerk off via e-mail?
I get why you got earthquakes wrong. You took a random selection of samples, and your sample size was too small, and your small dataset was not representative.
I guess if I had chosen a small subset of athletes at random to measure, and projected these measures onto a larger group, I risk making the same mistake.
What I'm not getting:
- What's all this about "the truth of all humanity"? The two metrics I made for a group/event are only intended to be interpreted for that group/event. I chose the parameters I chose, specifically to look for performance effects at the top. What happens with the rest of humanity below the top is another story with a different ending.
I didn't "(interpret) the presence or absence of a doping effect from a small number of performances" but "failed to observe a large performance improvement for the best performances of one of the groups I looked at", and asked how that can be explained in the EPO-era, given the many assumptions and promises often claimed for EPO.
a shill by any other name wrote:
rekrunner wrote:... I used a rather complete dataset.
Maybe so, but "complete" isn't remotely the same as statistically significant. In my counter example it took > 100,000 data points to disprove a false interpretation derived from > 50,000 data points. Here you are using a few dozen data points and presuming (because they capture "all" the top performances) that they represent the truth of all humanity. Big stretch... VERY big...
(I'm surprised you don't seem to "get" this, given the evident understanding conveyed in your first summary)
Racist? What race is North Africa? "5-continents" is pretty much all races.
1) I thought we fixed that. "Athletes from everywhere cheat, from 29-57% according to a recent Tuebingen survey".
2) I never assumed or concluded "Africans are genetically superior".
I also showed "non-Africans failed to improve massively as soon as epo became available".
"Athletes from Third World corrupt nations with no testing are more likely to cheat." And you called my assumptions "racist"?
Are you suggesting that none of the top athletes from "5 continents" took EPO during the EPO-era?
That's the opposite of what we saw in cycling.
Coevett wrote:
Begin with two ideological and racist assumptions :
1/ Everybody cheats.
2/ Africans are genetically superior.
Use psuedo-scientific pretence of being experts in statistical analysis to show what was already obvious- Africans improved massively as soon as epo became available.
Use assumptions 1 + 2 and psuedo-scientific pretence to prove assumptions 1 + 2 and ignore obvious explanation that athletes from Third World corrupt nations with no testing are more likely to cheat.
Why are you asking me?
Not everybody dopes but u2r dopes wrote:
Are Gary and Casual Observer actually different people? I somewhat doubt it, but if you are, couldn't you just have this mutal two way jerk off via e-mail?
0/10. What a wiseass comment! (imagine that...wiseass comments on LRC ?). There's been good & insightful information discussed here by rekrunner, casual observer, Coevett, amateur & others...and you come in here and throw a stupid wiseass crack like that. Is this subject matter too complicated for you to offer anything of value? Maybe college classses got out early and you have nothing better to do? ?
Not everybody dopes but u2r dopes wrote:
Are Gary and Casual Observer actually different people? I somewhat doubt it, but if you are, couldn't you just have this mutal two way jerk off via e-mail?
Rocket Fuel Rick wrote:
casual obsever wrote:
It also worsens the statistics. A lot. And potentially skews them when the person who wants to make a point chooses the sub-groups, like you did here.
casual obsever wrote:
But if you consider that the best performances post-1990 are EPO-aided, then you need to consider that the pre-1990 performances are also enhanced by the then basically undetectable (legal and illegal) methods.
Otherwise you'd postulate that the sport was clean pre-1990. We know better.
casual obsever wrote:
The question then becomes: how much - possibly gender- and distance-specific - helps (at first unlimited) EPO and controlled but more powerful steroids vs. unlimited blood transfusions and basically unlimited but primitive steroids. And yet, each time somebody brings this up, you argue against it.
casual obsever wrote:
On the other hand, if you were to think that runners are too honest to cheat even without noticeable risk of detection (and not even use blood transfusions while allowed), then this method were absolutely pointless. Then we can only look at the athletes with positive tests, and whose doping history was proven by admissions and witnesses. At the top, those are 0, as nobody talks, and everyone pretends they only doped once.
casual obsever wrote:
Finally, let's look at the numbers above. Why would the longer distances benefit more? Ok, the marathon was underdeveloped in the 80s, and now there is more money, so we can't really compare that.
But do 5000 m runners earn more than 800 m runners? Doubt it. More likely, the benefit from the steroids decreases with increasing distance, while blood doping methods become more important.
So long as we are sticking to what the record shows, Ceplak's PB is from 2002, and her EPO bust is from 2007. In 2007, she was running slower at 1:59 and 2:00.
Rocket Fuel Rick wrote:
And "doping busts" are clear cut and allows for the analysis of progression trends, PBs, competition record & performance post-ban for those who returned to competition. For example, Ceplak had a big 3 second jump in one year before she set her sizzling 1:55 PB. And when she came back from her ban, she was crap never running anything near her fastest times and not even coming close to breaking 2 again.
So, show me where Ceplak (was) using androgens? The record shows an EPO positive ....
rekrunner wrote:I didn't "(interpret) the presence or absence of a doping effect from a small number of performances" but "failed to observe a large performance improvement for the best performances of one of the groups I looked at", and asked how that can be explained in the EPO-era, given the many assumptions and promises often claimed for EPO.
I took a different 1 euro coin, out of the SAME pocket, mind you... and just produced the following series of results: H H T T T T H H T H H H H H H H T H H .
It seems like THIS coin tends to produce long series of heads, and also yields heads 70% of the time. Now I can do one of two things with this newly developed knowledge about my two coins:
(1) I can go into a bar, confident of my two coins' abilities and trick some guy into betting against me, when I *know* my coins are unfair. Or,
(2) I can recognize that both my coins are actually most probably fair, and will yield about 50% of each heads and tails, in the long run, and NOT throw away my money in the bar.
Here, you've walked into the wild west LRC saloon, threw your coin (race stats) down on the table and challenged us all to a bet about doping, believing in your heart your coin contains some hidden wisdom not known to the rest of us. I defy you! Your coin does NOT contain unfair hidden wisdom.
(All in fun of course; I do admire the effort you took to work out the stats and think it all through, I simply don't agree that any conclusions are warranted)
rekrunner wrote:
What do you suggest I skewed? Would you chose different sub-groups?
rekrunner wrote:
Did *I* consider the best post-1990 performances EPO-aided?
rekrunner wrote:
Fine. How do we get answers to the questions, without answering them ourselves with new assumptions? Were the *best* pre-1990 times a result of "unlimited blood transfusions and basically unlimited but primitive steroids"? Are the *best* post-1990 times a result of "(at first unlimited) EPO and controlled but more powerful steroids"?
