How to Lie With Statistics Post Hoc Rides Again
How to Lie with Statistics, by Huff
Monday April 5, 2021
Darrell Huff and his popular 1954 book have received criticism. Huff was incorrect about cigarettes and cancer, but he did warn us that people are biased. In 142 breezy pages, the book manages to cover quite a lot, with perchance moderate levels of prejudice for its time.
- Chapter 1: "The Sample with the Built-in Bias" (sampling)
- Chapter 2: "The Well-Called Average" (mean vs. median)
- Chapter three: "The Trivial Figures That Are Non There" (variance etc.)
- Chapter iv: "Much Ado well-nigh Practically Nil" (small differences and measurement uncertainty)
- Chapter 5: "The Gee-Whiz Graph" (funny y-axes)
- Affiliate half dozen: "The One-Dimensional Picture" (dimensional baloney of comparisons)
- Chapter vii: "The Semiattached Effigy" (non-sequitur logic)
- Affiliate 8: "Mail service Hoc Rides Again" (correlation ≠ causation)
- Affiliate ix: "How to Statistculate" ("statistically dispense"—nonsense math)
- Chapter 10: "How to Talk Back to a Statistic" (enquire questions)
- "Who says so?"
- "How does he know?"
- "What'south missing?"
- "Did somebody change the subject?"
- "Does it brand sense?"
"There are at least three levels of sampling involved. Dr. Kinsey's samples of the population (one level) are far from random ones and may non exist especially representative, simply they are enormous samples by comparing with anything washed in his field before and his figures must be accustomed as revealing and important if not necessarily on the olfactory organ. It is perchance more important to call up that any questionnaire is only a sample (another level) of the possible questions and that the answer the lady gives is no more than a sample (tertiary level) of her attitudes and experiences on each question." (page 23)
"Some of the strongest feeling against public-opinion polls is found in liberal or left-wing circles, where it is rather unremarkably believed that polls are generally rigged. Behind this view is the fact that poll results and then often fail to square with the opinions and desires of those whose thinking is not in the conservative direction. Polls, they point out, seem to elect Republicans fifty-fifty when voters before long thereafter practice otherwise." (folio 26)
Interesting historical perspective; modernistic concerns about polls are not necessarily new issues.
"You lot will also learn if yous read back into the tables that the figure is based on a sample of such size that there are 19 chances out of 20 that the estimate—$3,107 earlier information technology was rounded—is correct inside a margin of $59 plus or minus." (pages 35-36)
I don't know a confidence interval technique that will give you quite that interpretation; this could be erroneous.
"Caste of significance" (p-value) is presented on page 42 equally a thing that is usually hidden from the public, which may take been (and may still oftentimes be) the case. These days I hear more nearly problems with significance tests than demands for them.
"It is dangerous to mention any subject having high emotional content without hastily proverb where you are for or agin it." (page 46, amusing spelling in original)
"The Procrustean Statistic" (graphic folio 43)
"In somewhat the aforementioned fashion those lilliputian figures [reporting variance] that are missing from what are called "Gessell's norms" take produced pain in papas and mamas. Let a parent read, as many have washed in such places every bit Sunday rotogravure sections, that "a child" learns to sit down cock at the age of so many months and he thinks at in one case of his own kid. Let his kid fail to sit by the specified age and the parent must conclude that his offspring is "retarded" or "subnormal" or something equally invidious. Since half the children are leap to fail to sit by the time mentioned, a expert many parents are made unhappy. Of course, speaking mathematically, this happiness is balanced by the joy of the other l per cent of parents in discovering that their children are "advanced." But harm can come of the efforts of the unhappy parents to force their children to accommodate to the norms and thus be backward no longer." (pages 44-45)
"Newsweek once showed how "U. S. Old Folks Abound Older" by means of a chart on which appeared two male person figures, one representing the 68.2-year life expectancy of today, the other the 34-year life expectancy of 1879-1889.'"
Hither Huff is complaining that the person twice as tall appears eight times equally massive, only in that location are other issues with interpreting historical life expectancy...
"Who knows what germ causes colds, especially since it probably isn't a germ at all?" (folio 75)
What does he think causes colds? They knew most viruses in the 50s, didn't they?
"Let united states of america say that during a menstruation in which race prejudice is growing y'all are employed to "prove" otherwise. It is not a difficult assignment. Set up up a poll or, better yet, accept the polling washed for yous by an organization of good reputation. Ask that usual cross section of the population if they think blacks have as good a chance equally white people to go jobs. Repeat your polling at intervals so that you volition have a trend to report.
"Princeton's Office of Public Stance Research tested this question once. What turned upwards is interesting that things, peculiarly in opinion polls, are not always what they seem. Each person who was asked the question virtually jobs was too asked some questions designed to discover if he was strongly prejudiced against blacks. It turned out that people most strongly prejudiced were well-nigh likely to answer Aye to the question about job opportunities. (Information technology worked out that about ii-thirds of those who were sympathetic toward blacks did non think the black had as skilful a chance at a task equally a white person did, and well-nigh two-thirds of those showing prejudice said that blacks were getting equally adept breaks equally whites.) It was pretty axiomatic that from this poll you would learn very little about employment conditions for blacks, although you might learn some interesting things well-nigh a human'due south racial attitudes." (pages 75-76)
A couple hit dehumanizing "blacks" vs. "white people" phrasings here. The reportage of veiled (?) racism in survey responses notwithstanding seems relevant today.
"A noncombatant population includes infants, the one-time, and the ill, all of whom have a higher expiry rate wherever they are." (page 85)
It'southward incidental to the point of the text here, merely it's interesting to encounter a reference to infants having a high death rate. I recollect this sounds out of place, today. When Huff was born in 1913, something similar ten pct of babies died earlier age five. (historical life expectancy)
"Keep in mind that a correlation may be existent and based on real crusade and effect—and all the same be almost worthless in determining activity in any single case." (page 93)
I read this equally pointing to variability. Was he already thinking about smoking?
"But arbitrarily rejecting statistical methods makes no sense either. That is like refusing to read because writers sometimes utilise words to hide facts and relationships rather than to reveal them." (folio 121)
"I'll face up to the serious purpose that I like to remember lurks only beneath the surface of this volume: explaining how to look a phony statistic in the eye and face information technology down; and no less of import, how to recognize sound and usable data in that wilderness of fraud to which the previous chapters have been largely devoted." (page 122)
"You may be familiar with the Rudolf Flesch readability formula. Information technology purports to measure out how piece of cake a piece of prose is to read, past such unproblematic and objective items as length of words and sentences. Like all devices for reducing the imponderable to a number and substituting arithmetic for judgment, information technology is an appealing thought." (page 137)
I'm interested in the readability stuff, just it also strikes me that the last sentence there is quite a skilful joke:
"Like all devices for reducing the imponderable to a number and substituting arithmetic for judgment, it is an highly-seasoned idea."
That could be an epigraph!
Source: https://planspace.org/20210405-how_to_lie_with_statistics_by_huff/
0 Response to "How to Lie With Statistics Post Hoc Rides Again"
Post a Comment