wtfunny wrote:
you want to pretend an anonymous p̶o̶l̶l̶ "troll" on the site that doesn't even disallow repeated voting bears some kind of weight?
The poll did disallow repeated voting. If you go to the individual tabs page and click on "About/Methods" It says the following (check out the "data integrity" section):
Poll Methods
Data were collected in three waves of polls shown to LetsRun.com readers. Respondents were shown a list of athletes, their nationality, and their top accomplishments and asked to choose "Clean" or "Dirty." We chose to force respondents to select an answer instead of providing an "I'm not sure option." We acknowledge that people's perceptions of doping are always shrouded with uncertainty, however we are more interested in determining which way people leaned than measuring their uncertainty. To that end, we find people's Yes/No answers more interesting than those which account for uncertainty, which of course exists.
Is this a scientific poll? Does it mean anything?
No, this is not a scientific poll (i.e., one that provides an unbiased estimate of a population's view on something with a quantifiable margin of error). The respondents are not representative of the general population, but only of LetsRun.com readers who chose to vote. The poll tells us about LRC readers' perceptions of who achieved their accomplishments while clean. Our aim was not to determine whether someone doped, but if people perceive someone to have doped.
Demographic Data
Respondents were given the option to provide demographic data. Country data were augmented by matching respondents' IP addresses to a country using a geo-location IP database. This procedure allowed us to identify countries for over 99% of respondents. The frequency of which visitors read LetsRun.com was measured using a Google Analytics cookie which counts sessions by device.
Data Integrity
We aimed to allow readers to vote only once and to ensure all votes actually came from a human being. Readers were only allowed to submit one vote from a single IP address and/or browser. We threw out votes which were from spam-bots.
Statistical Tests
We provide statistical tests to assess whether or not the observed differences in perceptions across groups are likely to be real or a result of random variation. These tests are only applicable to LRC readers. Furthermore, the tests assume those who voted are a random sample of all LRC readers, which is an un-tested assumption. Nevertheless, the test results provide some insight into the likelihood of a perception of doping varying across a group in a systematic rather than random way. Two proportion z-tests were used to compare across dichotomous groups (e.g., gender). Chi-squared tests were used to test associations across multiple sub-groups (e.g., race/ethnicity, age).