In any case, there was not a significant difference between the minimal M1 and M2 estimates, except (as they noted) when the sample size is really small. 18% (and also 15% for men and 22% for women), were just the summary figures for 2011, but not 2013, which were 15%, 17% and 12%, respectively -- much lower for the women, and slightly higher for the men. It's also fair to say that in-competition testing from 2011 and 2013 Championships would not detect out of season doping, so for this (and other) reasons, 18% should be interpreted with these limitations in mind. Having said that, the results from 2001-2009 (pre-2009, and in and out of competition), from 2011, and from 2013 were all similar, around 15-20%. Aragon, I'm curious about your thoughts, if any, on this topic: getting back to interpreting the Tuebingen study, which used the Unrelated Question Model (UQM) to guarantee anonymity, have you seen subsequent studies that compare that survey method (UQM) with other survey questionnaire techniques (e.g. SSC), casting doubt on UQM for sensitive questions, and its inability to estimate survey non-compliance? In a different study on a different population, they found "A large discrepancy was observed in prevalence estimates for PED using the UQM (58.4%) and SSC (19.8%)", suggesting that even the conservative lower bound estimate from the Tuebingen study might still be too high, being based on several subjective assumptions of likelihood that ignore the psychology of athletes, clean and dirty, being properly motivated to comply with survey questionnaires. In the Appendix of the Tuebingen study, they briefly mention Single Sample Count (SSC), and give the raw data in Table 6, promising "A more comprehensive discussion of the SSC methods for adjusting for noncompliance, together with the SSC-derived and noncompliance-corrected doping estimations from these fi eld tests, will be presented in a subsequent publication." As far as I know, it has not been published yet. I had asked, and remain curious, for "casual obsever's" thoughts on this same topic, but he has not responded to previous inquiries. Further References: Comparison of UQM Method with SSC, Social Projection, and Network Scale Up Methods "A potential inflating effect in estimation models: Cautionary evidence from comparing performance enhancing drug and herbal hormonal supplement use estimates", Ricky A. James, Tamás Nepusz, Declan P. Naughton, Andrea Petróczi, "Psychology of Sport and Exercise" 14 (2013) 84-96, doi: 10.1016/j.psychsport.2012.08.003 A deeper description of Single Sample Count: "Estimating the Prevalence of Socially Sensitive Behaviours: Attributing Guilty and Innocent Noncompliance with the Single Sample Count Method", Nepusz, T., Petróczi, A., Naughton, D., Epton, T., & Norman, P., "Psychological Methods", 19, 334-355, doi: 10.1037/a0034961
Aragon wrote:
Quickly browsed through the issue and looked for your source, so I presume you are referring to the 2011 paper which indeed has also the figure of 18 % but which has some data from 2001 so it is debatable how relevant its implications are today.
The source for the 18 % figure in the link above is the recent paper detailing the 2011 and 2013 games ("Only last month, a study estimated that 18 per cent of endurance athletes were guilty of blood doping at the 2011 and 2013 World Athletics Championships") which has no "minimum estimate" claim as far as I remember.