All of those risks and flaws of statistics are well-known and there are methods to deal with or account for them that are employed in any good study. The problem is that these uncertainties are often not mentioned, or highlighted, in most reports, at least not to the public. The flaws in statistical evidence include risks of selection biases, or massaging of the information to present a particular point of view, or people just downright lying to the survey in large enough numbers to skew the result -- in short, the risks are entirely down to human errors. Are these not just as prevalent, if not more so, in anecdotal evidence where people are less likely to try to account for such errors?