![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
I got this (from the library) after ![[livejournal.com profile]](https://www.dreamwidth.org/img/external/lj-userinfo.gif)
rivka included it in her January books post. As she says, it's relatively slight: I read it this afternoon in between games of nethack and random pacing. If you've studied statistics in any detail, you may think "yeah, yeah, we know all this"; if you've studied statistics in detail, you're not the intended audience for this book.
The author deliberately doesn't assume much mathematical sophistication, but he does point out the dangers of innumeracy. He starts with what he calls "the worst—that is, the most inaccurate—social statistic ever." When he saw it in a dissertation proposal, he assumed the student must have mis-transcribed it; in fact, the student had taken it word-for-word from a journal article. The statistic in question was the claim that "Every year since 1950, the number of American children gunned down has doubled." Best does the arithmetic to show that this is not only false, but so obviously impossible that the journal editor should have caught it without needing to check references: start with one death in 1950, and powers of 2 take it past a million by 1970, and past the entire population of earth by 1983. The source of this mistake is a careless rewrite; the original statement was that "The number of American children killed each year by guns has doubled since 1950." (The author notes that this is less startling than the organization presenting it was suggesting, given that the U.S. population had increased by 70% in that timespan.)
The book's goal is to get readers to be more careful about statistics, including noticing when they don't make sense (for example, "35 active serial killers" and "4,000-5,000 victims per year" the latter number derived from a misunderstanding of police reports that the relationship between killer and victim was "unknown," meaning that when the initial report was written, the police don't know what, if any, relationship there had been, not that the two were necessarily strangers, nor yet that all killings by strangers involve serial killers); when unlike things are being compared; and the various motivations different people and groups may have for shading things.
The author says that when a social problem is first talked about, the media are likely to insist on some kind of numbers, and having no better basis, activists sometimes guess. Unfortunately, the guesses are not only printed, but believed and repeated, to the point that if, some time later, the subject is getting attention and there are funds for a study, numbers from research may be dismissed because "everyone knows" that the "actual" numbers are whatever scientific wild-assed guess someone gave a reporter. He argues that such things as homelessness are "social problems" in two senses, both that they are problems arising from a societal context, and that the ways that they are noticed and treated as problems is itself a social process, and one of the main drivers for the presentation of statistics to the general audience.
One useful idea in this book is the "dark figure", a term drawn from criminology that refers to the percentage of crimes (or any other event you're studying) that don't turn up in the statistics. There are patterns in how the dark figures change: when something is first talked about, it's all or almost all dark figure. Over time, that tends to shrink, but how far it shrinks varies. (Many minor crimes remain in the dark figures, but criminologists have reason to believe that nearly all homicides are reported.)
I'm glad I read this, and glad that I borrowed a library copy to do so.
rivka included it in her January books post. As she says, it's relatively slight: I read it this afternoon in between games of nethack and random pacing. If you've studied statistics in any detail, you may think "yeah, yeah, we know all this"; if you've studied statistics in detail, you're not the intended audience for this book.The author deliberately doesn't assume much mathematical sophistication, but he does point out the dangers of innumeracy. He starts with what he calls "the worst—that is, the most inaccurate—social statistic ever." When he saw it in a dissertation proposal, he assumed the student must have mis-transcribed it; in fact, the student had taken it word-for-word from a journal article. The statistic in question was the claim that "Every year since 1950, the number of American children gunned down has doubled." Best does the arithmetic to show that this is not only false, but so obviously impossible that the journal editor should have caught it without needing to check references: start with one death in 1950, and powers of 2 take it past a million by 1970, and past the entire population of earth by 1983. The source of this mistake is a careless rewrite; the original statement was that "The number of American children killed each year by guns has doubled since 1950." (The author notes that this is less startling than the organization presenting it was suggesting, given that the U.S. population had increased by 70% in that timespan.)
The book's goal is to get readers to be more careful about statistics, including noticing when they don't make sense (for example, "35 active serial killers" and "4,000-5,000 victims per year" the latter number derived from a misunderstanding of police reports that the relationship between killer and victim was "unknown," meaning that when the initial report was written, the police don't know what, if any, relationship there had been, not that the two were necessarily strangers, nor yet that all killings by strangers involve serial killers); when unlike things are being compared; and the various motivations different people and groups may have for shading things.
The author says that when a social problem is first talked about, the media are likely to insist on some kind of numbers, and having no better basis, activists sometimes guess. Unfortunately, the guesses are not only printed, but believed and repeated, to the point that if, some time later, the subject is getting attention and there are funds for a study, numbers from research may be dismissed because "everyone knows" that the "actual" numbers are whatever scientific wild-assed guess someone gave a reporter. He argues that such things as homelessness are "social problems" in two senses, both that they are problems arising from a societal context, and that the ways that they are noticed and treated as problems is itself a social process, and one of the main drivers for the presentation of statistics to the general audience.
One useful idea in this book is the "dark figure", a term drawn from criminology that refers to the percentage of crimes (or any other event you're studying) that don't turn up in the statistics. There are patterns in how the dark figures change: when something is first talked about, it's all or almost all dark figure. Over time, that tends to shrink, but how far it shrinks varies. (Many minor crimes remain in the dark figures, but criminologists have reason to believe that nearly all homicides are reported.)
I'm glad I read this, and glad that I borrowed a library copy to do so.
Tags:
![[identity profile]](https://www.dreamwidth.org/img/silk/identity/openid.png){kind=small}
From:
no subject
From:
no subject
I'm dubious every time I read a statistic like "A woman over 40 is more likely to be hit by a meteor than get married." How many people are hit by a meteor each year, really?
From:
no subject
One example of a questionable statistic that does get quoted is "you are more likely to die from asteroid impact than in a plane crash". The rationale behind this is that an asteroid impact big enough to kill a billion people is expected to happen roughly once every million years. In other words, there is a one in one million chance in any given year that a billion people will die from asteroid impacts, which averages out to a thousand people dying a year in this way - more, indeed, than die in aviation accidents.
Except that this is not in any way a meaningful 'average'. It is the product of an exceptionally unlikely event and the very severe outcome of that event. With a very high probability, the expected number of people dying from asteroid impact each year is zero, whereas we can expect a few hundred people to be killed in plane crashes annually. So-called averages that only become meaningful over time-scales hundreds of times longer than human history can hardly be applied to day-to-day life.