For more information
Call 1-800-351-0222
Call 1-800-351-0222
"HOW TO LIE WITH STATISTICS" (DARRELL HUFF).
Term Paper ID:20184
Get This Paper Free! or
Essay Subject:
Summary of work on biases, averaging, graphs, figures & their abuse.... More...
7 Pages / 1575 Words
2 sources, 16 Citations,
APA Format
$28.00
More Papers on This Topic
Paper Abstract:
Summary of work on biases, averaging, graphs, figures & their abuse.
Paper Introduction:
The purpose of this research is to examine the nonfiction book How to Lie with Statistics by Darrell Huff. The plan of the research will be to set forth the main ideas of the book as a chapter-by-chapter summary of the important ideas, including examples of misleading uses of statistics of the type presented.
The Sample with the Built-In Bias. Statistical measurement begins with assembling a credible sample of respondents on which to base conclusions that can be generalized from the sample to the population. There are two factors to consider: whether the respondents are truly representative of what is being measured, and whether the respondents tell the truth when they are asked questions. Both features of a sample may contribute to bias in the reporting of information. Huff cites a survey in which voters in 1936 were asked over the phone predict election
Text of the Paper:
The entire text of the paper is shown below. However, the text is somewhat scrambled. We want to give you as much information as we possibly can about our papers and essays, but we cannot give them away for free. In the text below you will find that while disordered, many of the phrases are essentially intact. From this text you will be able to get a solid sense of the writing style, the concepts addressed, and the sources used in the research paper.
The Semi-attached Figures. The range of thesample results, also called their deviation from the average, is alsosignificant. 39). to say nothingof . If the range or deviation is not reported, that, too, issignificant, and suggests that any average may be misleading. This chapter explains the logical fallacy, ifB follows A, then A caused B. . Huff adds, however, that numbers--graphs,decimals, fractions, percentages, statistics--often lend credibility whereit is unwarranted. . The plan of the research will be toset forth the main ideas of the book as a chapter-by-chapter summary of theimportant ideas, including examples of misleading uses of statistics of thetype presented. Huff declares that "whatever anintelligence test measures it is not quite the same thing as we usuallymean by intelligence. The deceptive thing about the little figure that is not there is that its absence so often goes unnoticed. 67). The Little Figures That Are Not There. Truly random samples are expensive andunwieldy, yet so-called stratified random sampling (p. Huff cites a survey inwhich voters in 1936 were asked over the phone predict election results,choosing Landon over Roosevelt not least because Republicans far more thanDemocrats in 1936 owned phones. Huff also uses the example of money bags, the larger amountbeing represented by a bag that is not only twice as high but twice as wideas the smaller bag. Huff says that manufacturers have used such graphs to showdramatic increases in union salaries over the years. The principal cautionhere is to look at unitary representations, as well as at what numbers arenot included in any graph being reviewed. This concluding chapter givesspecific guidelines for critical statistical thinking, "how to look a phonystatistic in the eye and face it down to look a phony statistic in the eyeand face it down . Huff citesstudies that declare average age in months for developmental skills ofinfants without qualifying declarations by means of the range. How to Talk Back to a Statistic. Huff cites thedifference between germ-killing experiments in controlled laboratoryconditions and the effect of antiseptics on the human body. The One-Dimensional Picture. . Graphs can be truncatedto show broader or narrower movements than fully representational graphswould show. 54). Huff, Darrell. This chapter focuses onartificial standards for which tests are created to measure against suchstandards, which are held to be but may not be universal. Huff's caution to skepticism in connectionwith child development is for consumers and dispensers of statisticalinformation in all areas. Chanceis one. (1982). 1986). It takes no account of social judgment . But more people are out of work, and those who do work put inlonger hours possibly for less money. Reportedly, the religious righthas been vigorously recruited to respond to such surveys. Ratio of comparison between two variables is just asimportant, on this view, as the fact of comparison itself, and the failureof a consistent ratio is the success of an intention to exaggerate to makea point on one hand or to deliberately mislead on the other. . The tendencyof American Cold Warriors to represent on graphs the sheer size and mass ofSoviet nuclear missiles as against the more compact American missilesconcealed the fact that tactically, directionally, and electronically themore compact versions were also superior weapons. Look with suspicion on anyversion in which the bars change their widths as well as their lengthswhile representing a single factor or in which they picturethreedimensional objects the volumes of which are not easy to compare"(1982, p. 122). This chapter covers what Huffrefers to as the statistically inadequate sample, which is used to report,usually for PR reasons, "results that are not indicative of anything"(1982, p. . 21) has built-inlimits to demographic representativeness (age, income, education). . For example, assume House A,with a family of four, uses 1 gallons of water a day, and House B, with afamily of three, uses 4 gallons of water a day. Other factors of assessment: The size of the sample,the reporting of correlation with (good) or without (bad) a measure oferror, and the base numbers and sample demographics behind any averagesreported, assumptions expressed or implied, the (attached or semi-attached)relationship between raw numbers and the conclusions drawn, and thepotential for numbers to be effective without being meaningful. how to recognize sound and usable data in thatwilderness of fraud" (1982, p. Base numbers and elements in a series are also oftenshifted, facilitating biased reporting or conclusions. . That is, whatever sample is used as the basis of astatistical report, it is too small to be meaningfully generalized.Ironically, however, the press and the popular culture latch onto theaverage number as if it were written in stone. This has broadimplications in controversial situations. Yet none of the numbers making up the average isrepresentative of the sample! But what getsreported? In the early 198 s, allHaitians were considered a high-risk group for AIDS by the Centers forDisease Control, but their eventual removal from the list was connected tothe emergence of information about transmission of the disease: "It turnsout that, in fact, many of the AIDS victims we were seeing in Haiti hadengaged in homosexual activity . How to Statisticulate. Tracking AIDS to the ends ofthe Earth. This technique finds anequivalent in reports that productivity of American workers is up. In the daze that follows the collision of statistics with the human mind,hardly anybody will notice the difference" (1982, p. 116). Who's included?" (1982, p. to sayimagination. There is a drought.There are different ways of reaching false conclusions about the rightmethod of water conservation: 1) each house should reduce water consumptionby 5 %, leaving House A with 5 gallons for four people and House B with2 gallons for three; 2) the neighborhood should reduce consumption by5 %, leaving (in this sample) Houses A and B with 125 gallons each. Its verbal equivalent is doublespeak, but Huffsays that incompetence or the need to simplify complex material for generalreaders is as much a problem with statistical measurement as perniciousmotives (1982, p. (1986, December). The fault is in the filtering-down process from the researcher through the sensational or ill-informed writer to the reader who fails to miss the figures that have disappeared in the process. He uses IQ andsimilar intelligence tests as examples. The Gee-Whiz Graph. Esquire. Thus comparingtwo persons whose IQs fall within the same range, making one "smarter" thanthe other may be meaningless. The term means using statistical material tomanipulate information. But the technique, aimedat members of Congress, had the desired effect of encouragingrepresentatives and senators to vote for increased defense spending. . This is a graphic device that uses oneicon to represent thousands, or even millions, in a comparative analysis.But Huff is once again skeptical, chiefly of pictures that are not drawn inratio. Questions of bias, conscious orunconscious, can be addressed by looking at the corporate or institutionalsource of a report. One of Huff's problems with averages issimple to state with an example: 1 +13+46=69. Thus results ofsuch 9 -number calls could be skewed. AsHuff says, "It is sometimes a substantial service simply to point out thata subject in controversy is not as open-and-shut as it has been made toseem" (1982, p. In fact,both B and A may be predicated of a third variable or more variables. It was an epidemiological red herring"(Allman, 1986, p. His purpose in this chapter is todistinguish between the terms in common use, such as arithmetical average(as above), the median (just as many above as below a figure). 57). Anotherexample would be the oft-cited correct cholesterol count, which has notonly confused consumers but also changed over the past few years, dependingon which study is released. 42). 35). The harm comes when fallacious relationshipssupport erroneous conclusions: Even today, Red Cross blood donors areobliged to report visits to Haiti. That, of course, is the secret of its success (1982, pp. 1 1). X antiseptic kills thousands of germs! It is in this waythat Huff comes to a discussion of the probable error and the standarderror (p. Such a problem would be typical of incomeoperating statements of many businesses in the same industry; Huff uses theexample of neighborhood incomes. The meaningful figures for employment are not reported.Huff reminds readers that a figure can be expressed in many ways: "You canexpress exactly the same fact by calling it a one per cent return on sales,a fifteen per cent return on investment, a ten-million-dollar profit" (p.82). "You must always keep that plus-or-minus inmind, even (or especially) when it is not stated" (p. It takes no account of social judgment . The purpose of this research is to examine the nonfiction book How toLie with Statistics by Darrell Huff. The average of these numbersis 69/3 = 23. Otherfactors of bias include interviewer incompetence, hostility of sampletargets to polling, and so on. "A bar chart is capable of deceit too. New York: W.W.Norton & Co.----------------------- 1 The Well-Chosen Average. In other words, averages, likesamples, may be biased, and it's good statistical business to be skepticalof how numbers are reported, which numbers aren't reported in a particularstudy, and so on. Whatthat conceals is that employment is down, which obliges remaining employedpersons to take up slack in production to maintain production levels.Assuming production levels remain constant, productivity output per workerincreases. So too would comparisons of MPG estimates ofcompact cars that fall within the same range; the difference betweenmileage on Car A and Car B may be statistically insignificant even thoughin a test Car A got 3 MPG and Car B 33 MPG, if the range for all such carsis 25-35 MPG. . Post Hoc Rides Again. diligence and emotional balance" (1982, p. Statistical measurement beginswith assembling a credible sample of respondents on which to baseconclusions that can be generalized from the sample to the population.There are two factors to consider: whether the respondents are trulyrepresentative of what is being measured, and whether the respondents tellthe truth when they are asked questions. . The semi-attached figure is that ofproductivity. Much Ado about Practically Nothing. But chieflyhe cautions, "When you see an average-pay figure, first ask: Average ofwhat? B may cause A instead of A causing B. . 45-6). or B and A may becorrelative, but neither may be conclusively shown to cause the other; Huffsays many medical studies fall into this pattern. Both features of a sample maycontribute to bias in the reporting of information. Another example: Callers are often askedto give opinions via 9 -numbers on TV. That is not necessarily the case. The Sample with the Built-In Bias. A good deal of the misunderstanding can be avoided if to the "norm" or average is added an indication of the range . . 74). "If you can't prove what you want toprove, demonstrate something else and pretend that they are the same thing. 55), which are numerical ranges between which numerical sampleresults may or may not fall. It neglects such important things as leadership andcreative imagination. What the probability willbe will differ from study to study, but a less than five percentprobability is considered statistically insignificant. Huff's principal points are that atruly representative sample "is one from which every source of bias hasbeen removed" (1982, p. References Allman, William F. Hundredth-place decimal units maybe graphically represented as integer units, allowing the creation ofgraphs with dramatically angled lines up or down. If a sample is truly of ameaningful size, for example, the report of the statistics will include thefigure for degree of significance, or the probability that the sample canbe used to generalize to the population (p. This chapter treats the wide use of graphicstatistical presentations skeptically. Huffcites elements that may influence the correlation between B and A. How to lie with statistics. 18); even so, he says, statistical reporters shouldbe skeptical of sample makeup.
If this paper is not what you are looking for, you can search again:
or
We can write a Custom Essay just for you.
Browse Essays by Subject