Abstract:
|
The American Statistical Association Statement on Professional Ethics (2000) is totally clear that undisclosed predetermination of estimation is a form of statistical misconduct. What is the consumer of statistical estimation to do when misconduct exists? In some important cases (e.g., jury decisions), competition among experts can lead naive jurors to at least unbiased, albeit inefficient, estimates. Here we suppose that statistical misconduct is one-sided. We consider how "anecdotal evidence" might be aggregated through an ideal political process to compete with biased experts. We give conditions under which the median of anecdotal evidence is an admissible estimator. We develop the "anecdotal" characteristic of estimation. We return to an idea first announced by Francis Galton in 1907 that majority-rule politics can be viewed as robust estimation. The historical context we explore is that of use of "experts" by cigarette companies in the late 19th, early 20th century to combat the "folk wisdom" that equated cigarettes with coffin nails. We shall focus on the activities of Cope's Tobacco of Liverpool, which attacked meddling evangelicals.
|