An approximate rant
| Gabriel |
In response to the first amendment case US v Stevens, Eugene Volokh comments on the majority’s statement that hunting videos are “several orders of magnitude” more popular than crush fetish videos. [This is important because the law at issue was motivated by fetish videos which are obscenity, but was drafted by Congress and used by prosecutors to also apply to non-obscene content]. Volokh presents the results of a reader poll he did recently showing that about a third of lawyers don’t know what “order of magnitude” means and on this he counsels that it might have been better for SCOTUS to say “more than a thousand times greater.” Given that a lot of lawyers don’t understand the phrase, Volokh is probably right for that audience. [It’s worth noting that Volokh himself is hardly innumerate, having earned a BA in math and computer science at the age of 15].
Similarly, the ASA Style Guide says that “order of magnitude” is too wordy and it’s better to use the word “ten.” I’m generally sensitive to the problem of wordiness (for instance, see Mark Twain’s fisking of Cooper) however the problem with the ASA’s advice is that “order of magnitude” does not mean “ten” but something more like “about ten.”
OK, so why not just say “about ten”? Because people will ignore the “about.” In a comment thread discussion I noted that even people who know better tend tend to reify point estimates. For instance, the advertising market is sensitive to measured changes in audience size that are well within the margin of error for audience surveys (Napoli 2003). What I like about the phrase “order of magnitude” is that unlike “ten,” or even “about ten,” the term makes it impossible to ignore the error bars.
The reason I agree with Volokh as regards to lawyers but disagree with ASA as regards to sociology is that lawyers can be forgiven for having priorities other than appreciating the uncertainty inherent in numbers but this should be a core issue for people who aspire to be scientists and if it isn’t then why the fuck do we bother making statistics a requirement at both the undergrad and grad level?